Nanoengineering of Structural Functional and Smart Materials

Nanoengineering of Structural,
Functional, and Smart Materials

Edited by
Mark J. Schulz, Ajit D. Kelkar,
and Mannur J. Sundaresan
Copyright 2006 by Taylor & Francis
Nanoengineering, SEM, nanofibers, nanoparticles,
Nanomaterials, Nanotubes
Library of Congress Cataloging-in-Publication Data
Preface
Need for the Book. In most areas of science and engineering, there is research
underway related to nanotechnology. However, the research is in different
disciplines and the basic and applied research is often not in step. The intent of this
book is therefore to connect science and technology under the umbrella of
nanoengineering in order to design and build practical and innovative materials and
devices from the nanoscale upward. Nanoengineering is fast becoming a cross-cutting
eld where chemists, physicists, medical doctors, engineers, business managers, and
environmentalists work together to improve society through nanotechnology. Nanoscale
materials such as nanotubes, nanowires, and nanobelts have extraordinary properties
and unique geometric features, but utilizing these properties at the nanoscale and
bringing these properties to the macroscale are very challenging problems. The
authors of the 24 chapters of the book explain these problems and have attempted to
develop well integrated coverages of the major areas where materials
nanotechnology has shown advances and where the potential to develop unique
structural, functional, and smart materials exists.
Structural materials are dened as load bearing and are designed mainly based
on mechanical properties. Examples where nanoscale materials can improve
mechanical properties include polymer and metallic materials reinforced with nano-
particles and thin lms to increase the surface hardnesses of the materials. Functional
materials are designed to have special properties, and are not primarily used for their
mechanical characteristics. Functional materials can have tailored or functionally
graded physical attributes such as electrical and thermal conductivity, magnetic
properties, gas storage, and thermoelectric properties, and sometimes graded
mechanical properties such as hardness. Nanoscale functional materials can be used
in high-tech applications including magnetic devices, electronics conducting ther-
moplastics, anisotropic polymer nanocomposites, surface coatings, biomaterials,
sensor materials, catalysts, polymers, gels, ceramics, thin lms, and membranes.
Smart or intelligent materials have sensing or actuation properties such as piezo-
electric or electrochemical transduction activities. Carbon nanotubes are smart mate-
rials because their electrochemical and elastic properties are coupled, and they have
higher theoretical actuation energy densities than existing smart materials.
Scope of the Book. Our goal is to provide readers with background in the various
areas of research that are needed to develop unique atomically precise multifunctional
materials that may be the strongest, lightest, and most versatile materials ever made.
The background needed to accomplish this encompasses synthetic chemistry, bio-
technology, self-organization, supramolecular self-assembly, nanophased particles,
lms and bers, chemical vapor deposition, oxide evaporation, and various
approaches to develop extraordinary strength, toughness, and net shape processing
of multifunctional materials and structures. In addition, molecular sensors, active
nanocomposites, thin lm skins, power generation,, high thermal and electrical
conductivities, and biomimetics are all discussed with the aim of optimizing material
systems to monitor their performance and maintain their integrity. These processes
may exploit the large elastic and transduction properties of carbon nanotube materials
for developing extraordinary multifunctional capabilities. Moreover, because the
nanotube structure is not limited to carbon, the benets of nanoscale inorganic
fullerene-like materials and nanotubes for developing multifunctional and polar
materials are also examined. Many elements and compounds are known to form
stable two-dimensional sheets, and hence create many exciting possibilities for
developing new types of nanoscale materials. Most materials that can be formed by
physical or chemical vapor deposition have the potential to form nanotubes, nano-
belts, nanowires, or some form of nanostructure, and many of these new materials
are discussed in this book. One goal of nanoscale research is to produce synthetic
analogic bionic materials that evolve their own nanostructures, sense and react to
their environments, self-monitor their conditions, and have super-elastic and self-
healing properties to provide enduring performance.
This book provides engineers and scientists the broad foundation needed to
attack barrier problems and produce high-payoff nanotechnology. As you will see,
this book contains quite a variety of research representing different approaches and
viewpoints about nanotechnology. Nanoengineering is a new eld, and this book
serves as a focal point and reference that can be used to conceptualize and design
new materials and systems. It was made possible by the generous contributions of
scientists from around the world, and presents state-of-the-art nanoscience and
nanotechnology including comments on future directions for research. The book
will help researchers, students, managers, those working in industry, and investors
understand where we are and where we are headed in the area of nanoscale,
nanophase, and nanostructured materials and systems. Many gures and detailed
descriptions of the synthesis, processing, and characterization of nanoscale materials
are included so that the book serves as a learning tool for nanoengineering, and so
that readers can reproduce the results presented. Problems are included at the end
of each chapter to test understanding of the concepts presented and to provoke further
investigations into the subject. This book is also meant to be used as a textbook for
graduate level nanotechnology courses. It is hoped that the book will inspire students
of all ages and disciplines to study nanotechnology and to think of different ways
to use it to help humanity.
Multifunctional Materials. This book also explores the multifunctionality that is
common in nature. Multifunctional materials have several important properties
simultaneously, such as a structural material that can also sense and actuate. Pres-
ently, no smart material is also a structural material. Multifunctionality is actually
a universal trait of biological materials and systems. Since the beginning of time,
biological materials and systems have been designed by nature from the smallest
components upward, and they have capabilities unmatched by man-made materials.
Therefore, it makes sense to integrate biomimetics and nanoengineering; biomimetics
provides the architecture for materials design, whereas nanoengineering essentially
provides the route to build materials starting on the atomic scale, as in nature.
Biologically inspired nanotechnology or bionanotechnology can be described as
the process of mimicking the chemical and evolutionary processes found in nature
to synthesize unique almost defect-free multifunctional material systems starting
from the nanoscale up. Bionanotechnology is becoming a new frontier in the devel-
opment of advanced biomedical, structural, and other materials. Bionanotechnology
is exemplied through biological materials constructed in layered anisotropic and
self-assembled designs that provide strength and toughness at the same time and
biological systems in which sensing and actuation are performed using millions of
identical parallel nerves and muscle bers. This architecture allows billions of bits
of sensory information to be processed in the neural, auditory, and visual systems
in an efcient hierarchical order and millions of identical micro-actuators to work
in harmony. In the book, initial concepts are discussed to mimic the basic functions
of nerves and muscles using nanoscale materials. These concepts may someday lead
to digitally controlled intelligent and enduring materials and structures.
Applications and Benets. The socio-economic benet of nanoengineering will be
ubiquitous and lead to improved safety, security, and standard of living throughout
the world. Future materials and structures will have vastly improved properties and
durability. Smart machines will control their own performance, preserve their integ-
rity, and partially self-repair when damaged, and when they are worn out or obsolete,
they will be programmed to demanufacture and be recycled into new machines.
Building without machining may be another outgrowth of nanoengineering.
Nanoengineering will produce new launch vehicles, lightweight agile aircraft, and
may allow the human exploration of space. Major areas of impact include future
space missions that will use hybrid nanocomposites to provide a wholesale reduction
in weight in space vehicle systems through material substitution, redesign, and
integration; autonomous recongurable structures will increase speeds, reduce fuel
consumption, reduce pollution, reduce noise, and provide lasting performance for
aircraft; intelligent materials will provide structural health and performance moni-
toring to prevent degradation and failure of structures in all types of critical appli-
cations; nanocoatings, llers, sprays, and lms will provide protection from abrasion,
EMI, heat, and provide articial skins for materials. Commercial applications of
nanocomposite materials potentially include all composite material products, brake
disks, turbine engine shrouds, composite bushings, brake parts, metallic composites,
smart materials, biosensing, and power harvesting. New applications will emerge as
our knowledge increases. Nanoengineering is also important in fuel cells where
functionalized nanotubes may store hydrogen safely for use in automobiles. Elec-
tronics, medicine, and computing are other areas where nanotechnology promises
advances. Indeed, our vision of nanoengineering is to obtain nanoscale control over
the synthesis of matter to build designer materials that can be used to solve the most
difcult scientic and medical problems that face humanity.
Outline of the Book. The book is organized into an introduction and three parts
that cover the major areas of focus in nanoscale materials development. The Intro-
duction to Nanoengineering chapter gives an insightful overview of where we are in
understanding nanoscale phenomena, and possible future directions for research.
Part 1 of the book is focused on Synthesis of Nanoscale Materials and contains beautiful
microscopy images of different synthetic materials, a comprehensive exposition of
the properties and synthesis of carbon nanotubes and bismuth nanowires, unique
methods of producing zinc oxide nanobelts, advanced methods for carbon nanotube
synthesis, synthesis of self-assembled nanodots, and a ball milling-annealing method
to synthesize boron nitride nanotubes.
Part 2 concentrates on Manufacturing Using Nanoscale Materials and includes
a technique for functionalizing nanoscale materials for material property improve-
ments, techniques for producing structural and metalceramic nanocomposites, the
use of low-cost carbon nanobers to form bers and lms, a comprehensive overview
of techniques for producing macroscopic bers from single-walled carbon nano-
tubes, a means of fabricating microdevices through self-assembled monolayers,
using nanotubes to improve the strength of polymers, properties and applications of
nanoscale intelligent materials, thermal properties of nanostructured polymers, and
pultruded nanocomposite materials.
Part 3 of the book focuses on Modeling of Nanoscale and Nanostructured
Materials. Nanomechanics and modeling of nanoscale particles and their vibration
properties are discussed, along with methods of continuum and atomistic modeling
of the nanoindentation of thin lms. Modeling of thin lm heterostructures, polar-
ization in nanotubes, uneven stress distribution in nanocrystalline metals, carbon
nanotube polymer composites, and multi-scale heat transport are also discussed.
In summary, this book provides a broad synopsis of the nanomaterials research
conducted in university and government labs. Because the size of the book is limited,
much of the important research in the eld could not be included. Readers are
therefore encouraged to use this book as a starting point from which to explore the
literature on nanotechnology, which is becoming more exciting every day.
The editors thank Cindy Carelli and Yulanda Croasdale of Taylor & Francis, and
Larry Schartman and Frank Gerner of the University of Cincinnati whose support
was instrumental in producing this book.
Editors
Mark J. Schulz is an associate professor of mechanical engineering and co-director
of the Cincinnati Smart Structures Bio-Nanotechnology Laboratory. This laboratory
integrates nanotechnology and biomimetics to develop new smart materials and
devices for structural and medical applications. The laboratory includes a nanotube
synthesis lab, a processing lab for nanoscale materials, and a smart structures and
devices lab. Research in the labs focuses on building structural neural systems using
continuous piezoceramic and carbon nanotube neurons and electronic logic circuits
for structural health monitoring, carbon nanotube array biosensors for cancer
diagnostics, active catheters for diagnostics and surgery, reinforcing polymers using
carbon nanotubes and nanobers, and developing wet and dry carbon nanober
hybrid actuators to control large structures. His contribution to the book is dedicated
to the memory of his parents, Jeanne and Joseph.
Ajit D. Kelkar is a professor of mechanical engineering and associate director of
the Center for Advanced Materials and Smart Structures (CAMSS) and the founding
member of Center for Composite Materials Research (CCMR) at North Carolina
Agricultural and Technical State University, Greensboro, NC. He is also a member
of the National Institute of Aerospace (NIA). His research interests include nite
element modeling, atomistic modeling, performance evaluation and modeling of thin
lms, and nanomechanics. Some of the projects he is presently involved with include
continuum and atomistic modeling of thin lms, nanoindentation studies of thin
lms, low-cost manufacturing of ceramic composites using a nanoparticle alumina
matrix, and the effects of alumina nanoparticles on the mechanical behaviors of
epoxy resins. In addition he is involved in the low-cost manufacturing of composite
materials, damage characterization of thin and thick composite laminates subjected
to low-velocity impact loading, fatigue behavior of textile composites, and nite
element modeling of woven and braided textile composites.
Mannur J. Sundaresan is an associate professor of mechanical engineering and
the director of the Intelligent Structures and Mechanisms Laboratory at North Carolina
Agricultural and Technical State University. This laboratory is dedicated to develop-
ing novel sensors, instrumentation, and signal processing techniques applicable to
smart structures and structural health monitoring. It also integrates the microme-
chanics of damage evolution in heterogeneous materials and structural health mon-
itoring techniques for the life prediction of such materials. He has worked in the
areas of micromechanics of damage evolution, development of novel processing
techniques for carboncarbon composite materials, and experimental mechanics.
Contributors
Christian V.D.R. Anderson
Department of Mechanical Engineering
University of Minnesota
Minneapolis, Minnesota
Debasish Banerjee
Department of Physics
Boston College
Chestnut Hill, Massachusetts
F. James Boerio
Department of Chemical and Materials
Engineering
University of Cincinnati
Cincinnati, Ohio
Donald W. Brenner
Department of Materials Science and
Engineering
North Carolina State University
Raleigh, North Carolina
Han Gi Chae
School of Polymer, Textile, and Fiber
Engineering
Georgia Institute of Technology
Atlanta, Georgia
Ying Chen
School of Physical Sciences and
Engineering
The Australian National University
Canberra, Australia
Richard O. Claus
Fiber and Electro-Optics Research
Center
Virginia Polytechnics Institute and State
University
Blacksburg, Virginia
Saurabh Datta
Smart Structures Bionanotechnology
Laboratory
Cincinnati, Ohio
Virginia A. Davis
Department of Chemical Engineering
Rice University
Houston, Texas
Mildred S. Dresselhaus
Massachusetts Institute of Technology
Cambridge, Massachusetts
Edward H. Glaessgen
Analytical and Computational Methods
Branch
NASA Langley Research Center
Hampton, Virginia
Rahul Gupta
North Carolina Agricultural and
Technical State University
Greensboro, North Carolina
Peng He
Laboratory
Cincinnati, Ohio
Yun-Yeo Heung
Laboratory
Cincinnati, Ohio
David Hui
University of New Orleans
New Orleans, Louisiana
Douglas Hurd
Laboratory
Cincinnati, Ohio
Farzana Hussain
Department of Mechanical and
Aerospace Engineering
Oklahoma State University
Tulsa, Oklahoma
Sachin Jain
Laboratory
Cincinnati, Ohio
Ado Jorio
Federal University of Minas Gerais
Belo Horizonte, Brasil
Inpil Kang
Laboratory
Cincinnati, Ohio
Ajit D. Kelkar
Goutham Kirkeria
Laboratory
Cincinnati, Ohio
Joseph H. Koo
University of Texas
Austin, Texas
Dhanjay Kumar
Satish Kumar
Engineering
Atlanta, Georgia
Young W. Kwon
and Energy Processes
Southern Illinois University
Carbondale, Illinois
Department of R&D
Inovenso Ltd. Co.
Istanbul,
Kin-Tak Lau
The Hong Kong Polytechnic University
Hong Kong, China
Hongbing Lu
Tulsa, Oklahoma
John F. Maguire
Air Force Research Laboratory
WrightPatterson Air Force Base,
Ohio
Hassan Mahfuz
Department of Ocean Engineering
Florida Atlantic University
Boca Raton, Florida
David B. Mast
Cincinnati, Ohio
Vincent Meunier
Computer Science and Mathematics
Division
Oak Ridge National Laboratory
Oak Ridge, Tennessee
Atul Miskin
Laboratory
Cincinnati, Ohio
Serge M. Nakhmanson
Suhasini Narasimhadevara
Laboratory
Cincinnati, Ohio
Jagdish Narayan
Department of Material Science and
Engineering
Marco Buongiorno Nardelli
Computer Science and Mathematics
Division
Oak Ridge National Laboratory
Oak Ridge, Tennessee
Sudhir Neralla
Department of Mechanical
Engineering
Gregory M. Odegard
Engineering
Michigan Technological University
Houghton, Michigan
Sri Laxmi Pammi
Laboratory
Cincinnati, Ohio
Matteo Pasquali
Department of Chemical Engineering
Rice University
Houston, Texas
Dawn R. Phillips
Lockheed Martin Space Operations
Hampton, Virginia
Louis A. Pilato
KAI, Inc.
Austin, Texas
Oded Rabin
Massachusetts Institute of
Technology
Cambridge, Massachusetts
Zhifeng Ren
Boston College
Chestnut Hill, Massachusetts
Samit Roy
Tulsa, Oklahoma
Erik Saether
Analytical and Computational Methods
Branch
Hampton, Virginia
Jagannathan Sankar
J. David Schall
Engineering
Mark J. Schulz
Laboratory
Cincinnati, Ohio
Vesselin N. Shanov
Laboratory
Cincinnati, Ohio
Donglu Shi
Nanoparticle Coating Laboratory
Cincinnati, Ohio
Vishal Shinde
Laboratory
Cincinnati, Ohio
Mannur J. Sundaresan
Lakshmi Supriya
Fiber and Electro-Optics Research
Center
Virginia Polytechnic Institute and State
University
Blacksburg, Virginia
Kumar K. Tamma
U.S. Army High Performance
Computing Research Center
Ashutosh Tiwari
Engineering
Tetsuya Uchida
Engineering
Atlanta, Georgia
Kalivarathan Vengadassalam
Tulsa, Oklahoma
Xudong Wang
School of Materials Science and
Engineering
Atlanta, Georgia
Zhong Lin Wang
School of Materials Science and
Engineering
Atlanta, Georgia
Cindy K. Waters
Jim S. Williams
School of Physical Sciences and
Engineering
The Australian National University
Canberra, Australia
Vesselin Yamakov
National Institute of Aerospace
Hampton, Virginia
Sergey Yarmolenko
Engineering
Contents
Chapter 1
Introduction to Nanoengineering
John F. Maguire and David B. Mast
PART 1 Synthesis of Nanoscale Materials
Chapter 2
Design of Nanostructured Materials
Debasish Banerjee, Jingyu Lao, and Zhifeng Ren
Chapter 3
Carbon Nanotubes and Bismuth Nanowires
Mildred S. Dresselhaus, Ado Jorio, and Oded Rabin
Chapter 4
Nanobelts and Nanowires of Functional Oxides
Xudong Wang and Zhong Lin Wang
Chapter 5
Advances in Chemical Vapor Deposition of Carbon Nanotubes
Vesselin N. Shanov, Atul Miskin, Sachin Jain, Peng He,
and Mark J. Schulz
Chapter 6
Self-Assembled Au Nanodots in a ZnO Matrix: A Novel Way to Enhance
Electrical and Optical Characteristics of ZnO Films
Ashutosh Tiwari and Jagdish Narayan
Chapter 7
Synthesis of Boron Nitride Nanotubes Using a Ball-Milling and Annealing
Method
Ying Chen and Jim S. Williams
PART 2 Manufacturing Using Nanoscale Materials
Chapter 8
Plasma Deposition of Ultra-Thin Functional Films on Nanoscale
Materials
Peng He and Donglu Shi
Chapter 9
Structural Nanocomposites
Hassan Mahfuz
Chapter 10
Synthesis and Characterization of Metal-Ceramic Thin-Film
Nanocomposites with Improved Mechanical Properties
Dhanjay Kumar, Jagannathan Sankar, and Jagdish Narayan
Chapter 11
Macroscopic Fibers of Single-Walled Carbon Nanotubes
Virginia A. Davis and Matteo Pasquali
Chapter 12
Carbon Nanober and Carbon Nanotube/Polymer Composite Fibers
and Films
Han Gi Chae, Tetsuya Uchida, and Satish Kumar
Chapter 13
Surface Patterning Using Self-Assembled Monolayers: A Bottom-Up
Approach to the Fabrication of Microdevices
Lakshmi Supriya and Richard O. Claus
Chapter 14
Enhancement of the Mechanical Strength of Polymer-Based
Composites Using Carbon Nanotubes
Kin-Tak Lau, Jagannathan Sankar, and David Hui
Chapter 15
Nanoscale Intelligent Materials and Structures
Yun Yeo-Heung, Inpil Kang, Sachin Jain, Atul Miskin,
Suhasini Narasimhadevara, Goutham Kirkeria, Vishal Shinde,
Sri Laxmi Pammi, Saurabh Datta, Peng He, Douglas Hurd,
Mark J. Schulz, Vesselin N. Shanov, Donglu Shi,
F. James Boerio, and Mannur J. Sundaresan
Chapter 16
Thermal Properties and Microstructures of Polymer
Nanostructured Materials
Joseph H. Koo and Louis A. Pilato
Chapter 17
Manufacturing, Mechanical Characterization, and Modeling of a Pultruded
Thermoplastic Nanocomposite
Samit Roy, Kalivarathan Vengadassalam,
Farzana Hussain, and Hongbing Lu
PART 3 Modeling of Nanoscale
and Nanostructured Materials
Chapter 18
Nanomechanics
Young W. Kwon
Chapter 19
Continuum and Atomistic Modeling of Thin Films Subjected
to Nanoindentation
J. David Schall, Donald W. Brenner, Ajit D. Kelkar,
and Rahul Gupta
Chapter 20
Synthesis, Optimization, and Characterization of AlN/TiN Thin Film
Heterostructures
Cindy K. Waters, Sergey Yarmolenko, Jagannathan Sankar,
Sudhir Neralla, and Ajit D. Kelkar
Chapter 21
Polarization in Nanotubes and Nanotubular Structures
Marco Buongiorno Nardelli, Serge M. Nakhmanson, and Vincent Meunier
Chapter 22
Multiscale Modeling of Stress Localization and Fracture in Nanocrystalline
Metallic Materials
Vesselin Yamakov, Dawn R. Phillips, Erik Saether, and Edward H. Glaessgen
Chapter 23
Modeling of Carbon Nanotube/Polymer Composites
Gregory M. Odegard
Chapter 24
Introduction to Nanoscale, Microscale, and Macroscale Heat Transport:
Characterization and Bridging of Space and Time Scales
Christianne V.D.R. Anderson and Kumar K. Tamma
1
Introduction
to Nanoengineering
John F. Maguire and David B. Mast
CONTENTS
1.1 Thermodynamic and Statistical Foundations of Small Systems
1.2 Definitions
1.2.1 Characterization and Metrology Needs
1.2.2 Computer Modeling Needs
1.3 Boundaries for Nanoscience and Technology
1.4 Some Final Thoughts
References
Nanoengineering offers the very real promise of a veritable cornucopia of enabling
new materials, devices, and products. Examples range from improved materials for
everyday uses such as self-cleaning paints and bathroom surfaces and deicing surface
treatments for aircraft and automobiles in northern climates to new forms of struc-
tural materials that might be stronger than steel yet lighter than Styrofoam. The
technology may also enable the development of adaptive soft materials like foams
and polymer composites that could enable fundamentally new sorts of products. For
example, imagine computers in which a CPU the size of a sugar cube has vastly
more computing power than all existing machines combined, or wallpapering a room
using paper that acts as a very-large-screen television. Similarly, work is under
way to produce very large mirrors from nanoengineered plastics so that huge,
lightweight plastic mirrors can be launched into space and used to see nearly the
beginning of time. Many major companies have recognized that this area of science
and technology holds the key to new products, processes, technologies, and medi-
cines, with multifaceted societal and economic benefits.
The above paragraph resonates with the kind of hyperbole that has become
the lingua franca of much of the nanoscience and technology (NST) discussed in
the popular press. Although many of these benefits may very well come true, and
some are already here, it would be exceedingly naive to expect that such major
innovations as quantum computing and advanced smart materials might be
developed without substantial scientific and technical breakthroughs on the one
hand, and concomitant major capital investment on the other.
Whereas some of the key enabling advances monumental discoveries such as
buckyballs, carbon nanotubes, and so forth have already been made, some are still
coming down the pipeline. In particular, it will be necessary to develop fundamentally
new paradigms in nanoengineering if the great promise of nanoscience and technology
is to translate into concrete societal benefits and the creation of wealth. It is of critical
importance, therefore, to quantify as fully and realistically as possible the scale and
scope of the scientific, engineering, and manufacturing challenges that must be met
to make even a rough order of magnitude estimate of the return on investment.
What Maxwell
1
once said about the field of classical thermodynamics is also appli-
cable to the development of the field of nanoengineering. Nanoengineering must meet
three essential criteria: there must be a solid scientific foundation, there must be clear
definitions, and there must be distinct boundaries. So, how about NST? How sound are
the foundations and where might they be strengthened? How good are the definitions
and standards and how might they be improved? And where indeed are the boundaries?
The remainder of this introduction will address these questions in the course of
providing a general and brief review of the background of NST. We should point
out that we make no effort to provide any kind of review of the field but simply
point out a few salient works. Our discussion also must focus on the non-bio aspects
of NST as related to the theme of this book.
In the broadest sense, NST represents the work of human minds expressed by
human hands to add value to and to create wealth from the natural resources of the
Earth. The subject is concerned with how to turn the very dirt of the Earth, the
metal ores and the crude oil, for example, into the automobiles, aircraft, aircraft
engines, computers, materials, and medicines of our advanced civilization. This cycle
requires the accumulated interdisciplinary knowledge of generations of scientists
and engineers and represents the real, or bedrock, knowledge environment on
which our economy is based and against which progress should be measured.
The essential basis for the ongoing NST revolution is the development of materials
and devices that operate over mesoscopic distance scales, where the material response
depends ultimately on the behavior of matter in what are called thermodynamically
small systems. The small is used here in a precise scientific sense that refers to a piece
of matter (or material system) where the size of the system is of the same order as some
relevant correlation length, [, such as the length scale related to interparticle interactions
or magnetic moment orientation. Behavior similar to that of thermodynamically small
granular systems often occurs near a critical point or phase transition or within a thin
interface between different materials, where the thickness of the interface is approxi-
mately equal to [. The importance of developing a deeper understanding of these types
of systems has long been recognized and as such has been demonstrated by the awarding
of several recent Nobel Prizes in Physics and Chemistry.
2
1.1 THERMODYNAMIC AND STATISTICAL
FOUNDATIONS OF SMALL SYSTEMS
In his now famous talk, Theres plenty of room at the bottom, Richard Feynman
3
essentially posed the nanomanufacturing problem in terms of a challenge for min-
iaturization. Feynman was very careful to dispose of the normal miniaturization
that had been ongoing, especially in the electronic industry from the early 1960s,
and made very clear that he was referring to the manipulation of matter over atomistic
distance scales. His foresight has captured the imagination of a generation of phys-
icists, engineers, and more recently chemists. It should not be forgotten, however,
that Professor Feynmans lifelong interest and particular scientific expertise was in
the area of statistical mechanics of dense many-particle systems. One of his earliest
achievements was to show that superfluidity in liquid helium was due to long-range
quantum correlation effects. It can be presumed that Feynman was very well aware
that a search for the ultimate in miniaturization would, of necessity, reveal a world
of new physical phenomena in which the macroscopic physics of our everyday
experience would require significant revision.
A number of early workers have made pivotal contributions to the foundation
of NST. Indeed, Sir John S. Rowlinson
4
has pointed out that understanding the basic
science of how forces between molecules result in the observable properties of
matter, especially soft matter at interfaces,
5
really represents a research effort of
almost 200 years in the making. However, there has been particularly significant
progress in our understanding of intermolecular forces and phase behavior over the
last 30 years with the award of the Nobel Prizes for the renormalization group theory
and the work in soft matter and polymers, as well as the discovery of nanoparticles.
For example, it is now well known that the correlation length is on the order of the
system size in three common situations (i.e., near a critical point, within an interface,
and in granular materials). Notice that it is the ratio of system size to the correlation
length that is important so that nano has nothing to do with a particular subdivision
of the meter, per se. In this regard, use of the term meso (in between) would seem
preferable to nano (dwarf). It is the ability to organize the mesoscopic structure of
matter and hold spatial coherence over macroscopic distance scales that lies at the
heart of nanoscience and technology.
These considerations have a number of examples. When machines become very
small, the van der Waals
6
forces become more significant, and the components tend
to undergo a jump to adhesion when they come close together.
7
Similarly, if materials
are sheared over nanoscopic distance scales in low dimensionality they do not wear
by stripping individual atoms from the surface but set up complex long-ranged vortex
patterns in which solid matter tends to flow like a cold plastic solid.
8
Even though one
might make a nanomachine, how would one lubricate it, and for how long would it
run? The alkanes that form the basis of everyday oils and lubricants crystallize into
two-dimensional structures at nanoscopic interfaces.
9
Although it is straightforward
to build simple computer models of molecules that look like turbine engines, gears,
or trucks, it must never be forgotten that matter over these distance scales obeys the
laws of statistical mechanics (the bridge between classical mechanics and quantum
mechanics). It is the free energy (perhaps local free energy) that matters, and if an
attempt were made to make such a molecular machine, it would certainly show a
tendency to spontaneously jump into reverse and, at worst, might even separate into
two or more liquid phases. Similarly, imagine that one deposits a feature that is 10
atoms wide (~2 nm) and 5 atoms thick using a metal deposited on a semiconductor
surface. Would such a structure tend to diffuse over the surface, would it remain
stable as deposited, or would it tend to rearrange into some other nanostructure?
Questions of thermodynamic and kinetic stability in nanoscopic structures are critically
important in this regard and will be ignored at our peril.
To achieve the promise of nano, it will be necessary to develop new nanoman-
ufacturing technologies that can actually produce materials and devices in large
quantities and at low cost in a reasonable time. Here it is important to recognize
that while it is possible at the moment to make nanoscopic structures and make
pictures (the famous IBM logo made from atoms is an excellent example), those
techniques that involve direct atomistic manipulation will be far too slow to produce
bulk products. If we could pick up and place an atom, perhaps with the tip of a
scanning microscope, at a rate of one per second, it would take ~10
23
s or longer
than the age of the universe to make a single mole of product. While this form of
nanoassembly might be appropriate for some applications involving atto-moles of
material, clearly, it will be necessary to manipulate phase transitions or directed
assembly techniques to arrange macroscopic matter in the required nanoscopic and
mesoscopic structures. The chemical physics of small nanoscopic and mesoscopic
systems is dominated by phase transitions that are largely driven by local entropic
contributions to the free energy. There has been a good deal of excellent theoretical
work in the chemical physics of small systems over the last 10 years, but little of it
has yet found its way into the current nano literature.
1.2 DEFINITIONS
The important point here is that before engaging in major real-world manufacturing
enterprises, it is absolutely essential that proper design tools be developed and tested.
The design tools in this case are the well-founded models (statistical physicsbased)
and measurement techniques that will generate the database on which reliable design
and manufacturing plans can be established. There are few areas of technical
endeavor in which recent fundamental theory and experiment bear so directly on
emerging manufacturing practice as in NST.
To produce these nanomaterials and devices in the required quantities, quality,
and cost, it will be necessary to do the following:
(a) Develop new nanomaterials characterization and metrology techniques
that probe the relevant structural property of matter over an appropriate
domain of energy and momentum.
(b) Explore theoretical and computer-modeling approaches that can help
rationalize observed behavior and response and, more importantly, predict
materials response in situations where direct measurement would be difficult
or impossible.
(c) Integrate the experimental and theoretical knowledge gathered in (a) and
(b) above and provide a seamless transition of this knowledge into the
engineering and nanomanufacturing environment using advanced tech-
niques such as computational methods in artificial intelligence.
10
It is, therefore, essential that methods be developed to predict and measure the
structure and properties of materials that are organized through directed assembly
techniques. These might be achieved by templating using a surface or near-surface
field or possibly by some combination of electric, magnetic (possibly of multipolar
symmetry), and flow fields. These and related approaches offer the possibility of
directing the nanoscopic and mesoscopic phase behavior and structure of matter in
ways that hold the promise of producing new forms of matter in quantities that will
be commercially viable.
1.2.1 CHARACTERIZATION AND METROLOGY NEEDS
There has been tremendous growth in the past 15 years in the development of tools
for the imaging and manipulation of nanometer-scale materials, even atoms. In truth,
the ability to image individual atoms with a Scanning Tunneling Microscope (STM)
was one of the defining developments that gave birth to the nanoscience and
nanoengineering age. The STM, and its cousin, the Atomic Force Microscope
(AFM), have since become the grandparents of whole families of different Scanning
Probe Microscopes (SPM) that are now known by their initials: LFM (Lateral Force),
SCM (Capacitance), MFM (Magnetic Force), and so forth. Also of importance is
the ability to manipulate nanometer-scale objects and even single atoms. The previ-
ously mentioned IBM logo spelled out by moving individual xenon atoms on a nickel
surface and the building of quantum corrals
11
for trapping atoms are well-known
examples of using an STM to manipulate single atoms. Equally impressive has
been the use of optical methods such as laser tweezers
12
for the manipulation of
biological samples such as DNA and optical traps
13
for the confinement and cooling
of atoms, for example in Bose-Einstein Condensation.
14
However, there are still
critical metrology requirements that need to be met and tools that need to be
developed before effective nanomanufacturing can take place. Some examples of
such requirements and tools are as follows:
Rapid detection of individual functional groups over large-scale surface
areas and interfaces. Spectroscopic techniques have long been of critical
importance in many areas of science and engineering, and it is to be
expected that the use of such techniques will play a central role in the
continuing development of NST. For example, consider Raman spectros-
copy, a powerful tool for the determination of the vibronic properties of
solids and liquids. The usefulness of Raman spectroscopy was further
extended to small and even single molecular groups with the use of Surface
Enhanced Raman Spectroscopy (SERS) techniques.
15
With SERS, these
molecular groups can be studied by attaching them to or near the surface of
nanometer-sized gold particles and using laser-induced surface plasmon
modes to greatly enhance the local Raman scattering. Unfortunately,
SERS is not a spatially scanning technique like the STM and AFM
previously discussed. For scanned optical spectroscopy at spatial resolu-
tions on the order of 50 nm, various near-field techniques such as a Near-
field Scanning Optical Microscope (NSOM) or total internal reflection
using a Solid Immersion Lens have been widely used. Work to develop
scanning, nanometer-scale Raman probes that will further reduce the size
of the scattering volume so that individual functional groups or moieties can
be investigated with spatial resolution on the order of 1.0 nm will be of key
importance to many areas of NST, both wet and dry. Recent work by
Novosky
16
using a scanned sharp metal tip, and Pettinger et al. and Mast,
17
using a STM, have begun the development of such a scanning nano-Raman
system. As well as providing the ultimate in surface spectroscopy, it will be
recognized that such technology would also allow, for example, the very
rapid mapping of genomes using massively parallel fiber-optic nanoprobes.
This will clearly be of central importance in the health-care industry.
A cartoon depicting the STM-Raman approach is shown in Figure 1.1, with
a photograph of such a system in the authors (DBM) lab shown in Figure 1.2.
Figure 1.3(a) shows the Raman spectrum of a thin sample of p-nitrobenzoic acid
taken with a conventional micro-Raman system, and Figures 1.3(b) and (c) show
the Raman spectra taken with the STM tip positioned over the sample and retracted
back from the sample.
Determination of local and interfacial mechanical properties using non-
contact and light-scattering measurements. Of critical importance for
many areas of NST is controlling the mechanical properties at the interface
of composite materials. Central to obtaining this desired control is having
the capability of accurately measuring these properties. For example, if
accurate Rayleigh-Brillouin studies could be conducted using near-field
interfacial scattering at this level of spatial resolution, it may well prove
FIGURE 1.1 Cartoon of combined Raman-STM System.
Scattered light
Spectrometer
Microscope objective
Incident light
STM tip
Nanostructured
material
FIGURE 1.2 (Color figure follows p. 12.) Photograph of miniature STM mounted on the
Raman microscope stage.
FIGURE 1.3(a) The Raman spectrum of the p-nitrobenzoic acid sample with the laser focused
directly onto the sample; the sample plane is perpendicular to the illumination/detection direction.
1177.8
1279.2
1286.9
1306.5
1320.4
1366.5
1385.2
1405.3
1438.7
1490.2
1515.8
1540.3
1596.6
1604.7
1627.2
1636.1
1129.1
1107.1
1350.9
10000
8000
6000
4000
2000
0
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
1000 1100 1200 1300 1400 1500 1600 1700
Raman shift (cm
1
)
Sample: p-nitrobenzoic acid
Accumulation time: 30 sec
Status: directly under microscope
(a)
possible to measure experimentally the local free energy as a function of
position (at nanometer resolution) through an interface. Such experimental
information is absolutely vital if we are to understand materials transport
and reaction in interfaces and thin films, including the cell wall.
Multiplexed sensor arrays for massively parallel detection. It is clear that
nanoengineered materials will be used in the near future to produce very
large structures. For example, if suitable nanoparticles are finely dispersed
in polymer films, it will be possible to engineer the microstructure such
that the films will conform to a particular shape on application of a stress.
Huge lightweight inflatable mirrors could be produced using this
approach. How does one measure such microstructure over meters and
even tens of meters in length? Although the single probe metrology tools
previously described are essential to characterize the initial development
of nanoengineered materials, they are many orders of magnitude too slow
for use on large systems. Tools will need to be developed that contain
arrays of large numbers of individual sensor probes for massively parallel
operation. In addition, software will be needed that will allow these
massively parallel tools to intelligently adapt their spatial resolutions for
the detection of defects at different spatial resolutions.
FIGURE 1.3(b) The Raman spectrum of the p-nitrobenzoic acid sample with the laser
focused directly onto the STM tip when the STM tip is close enough to the sample for actual
STM operation. The sample plane is parallel to the illumination/detection direction.
1200
0
20
40
60
80
1104.6
1349.8
1594.7
160
140
120
100
20
40
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
1000 1100 1300 1400 1500 1600 1700
Raman shift (cm
1
)
Status: STM mode in STM stage
(b)
1.2.2 COMPUTER MODELING NEEDS
On the modeling front, our understanding of the fundamental nature of intermole-
cular and interparticulate forces has increased considerably over the last 20 years.
It can now be said with some justification that the essential physics of simple dense
systems with rapid (~10
12
s) relaxations are reasonably well understood. This is,
however, not the case for complex molecular fluids or polymers. Here the relaxation
times can be relatively slow (>10
3
s), and the phenomena may be highly cooperative
and long range. Prior to the development of recent methods based on the application
of artificial intelligence in statistical mechanics,
18
such systems were quite outside
the realm of exact, albeit numeric, machine calculations. With the development of
new machine simulation techniques,
19
such systems are now amenable to more
rigorous treatment, and many interesting results will undoubtedly follow. The
coupling of artificial intelligence techniques with massively parallel machines
brings the solution of a number of realistic nanomaterial simulations within reach
if not yet quite within grasp. This should allow, for example, the first direct
simulations of nucleation and crystal growth from equilibrium along the melting
curve.
FIGURE 1.3(c) The Raman spectrum of the p-nitrobenzoic acid sample with the laser focused
directly onto the STM tip when the STM tip moved a fraction of a micron away from the
sample at the same point above the sample as in Figure 1.3(b). The sample plane is parallel
to the illumination/detection direction.
1000 1200 1300 1600
0
20
40
60
80
40
20
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
1100 1400 1500 1700
Raman shift (cm
1
)
Status: STM stage; 0.5 micron away from the tip
(c)
Similarly, early work has shown the potential importance of external fields on
surfaces and how the presence of a surface affects the conformation and adsorption
of individual molecules.
20
In addition, the mechanical properties and functionality
of advanced nanocomposite materials critically depend on the processes that produce
the mesoscale structure at the nanoparticlehost matrix interface. This is especially
true for granular nanomaterials.
Mesoscale simulations can be used to predict process dependence of properties
for granular nanomaterials and will enable electronic prototyping of new continuous
production processes to tailor and characterize the properties of nanomaterials. For
example, Figure 1.4 shows the results of a numerical simulation, of a two-component,
hard, spherical, granular material under the combined influence of gravity and a
sinusoidal vibration. This figure shows the positions of each particle in the ensemble
after a large number of vibration cycles and reveals the wide variation of order that
has developed.
21
Again, new methods will allow the study of larger ensembles where
the fully coupled nature of the phenomena, possible near phase transitions, will
undoubtedly enable serious investigation in systems ranging from catalysis to the
nature of protein interaction on cell walls or the formation of mesoscale entropically
driven structures.
FIGURE 1.4 Showing the results of a computer simulation on the thermodynamics of
fluidized powders investigating gravitational effects on mixing and segregation. This sim-
ulation is on a system comprising two dissimilar hard spheres under vibration. These
simulations show that the system forms an ordered two-component solid near the bottom
of the column, followed sequentially upward by a two-component fluid, an ordered single-
component solid, a single-component fluid, and finally a single-component gas. (Image
courtesy of L.V. Woodcock.)
1.3 BOUNDARIES FOR NANOSCIENCE
AND TECHNOLOGY
One of the very important issues that scientists and engineers need to keep in mind
is that, like the well-known proverb, All that glitters is not gold, not all that is
small is nano. Scientists and engineers, as individuals and as part of groups and
institutions, feel a strong pull to join in, to be a part of all exciting fields of endeavor;
this has been especially true for all things nano. The National Nanotechnology
Initiative, directed funding from NSF, DOD, and DOE, and national, regional, and
local nanoresearch centers have all contributed to the rush to add the word nano
to many well-established, existing fields of research and engineering. We must be
very selective with the use of nano lest we dilute its meaning and over-hype
expectations of the benefits of what nanoscience and nanoengineering can contrib-
ute to society. If we are not careful, this saturation of the field with inappropriate
uses of nano may likely result in a dilution of research funding below the amounts
necessary for success.
1.4 SOME FINAL THOUGHTS
The interdisciplinary nature of nanoscience technology and manufacturing is a
recurring theme. As evidence of the fruitfulness of the interdisciplinary approach,
we can cite the spectacularly successful work that has resulted at the interface
between chemistry and biology, called molecular biology. Work in this field has
given rise to the new discipline of molecular science, which has resulted in tre-
mendous advances in medicine, genetics, and agriculture. At the interface between
chemistry and physics there has also been a fusion of new ideas in an area variously
called materials chemistry, soft matter, and sometimes surface and interfacial
science. Here the focus is to understand the forces between atoms and molecules
in a dense medium and to use this knowledge to design new and useful forms of
matter. As Philip Ball writes in Made to Measure (quoted by Dr. Rita Colwell,
Director, National Science Foundation (NSF) in her address to the Materials
Research Society), We can make synthetic skin, blood, and bone. We can make
an information superhighway from glass. We can make materials that repair them-
selves, that swell and flex like muscles, that repel any ink or paint, and that capture
the energy of the sun. There is hardly a better example of how the old barriers
have crumbled. The power and momentum of NST lie in no small measure in its
porous boundary between physics, chemistry, and biology. Thus in the real world
of today we routinely turn chemists into physicists and physicists into materials
scientists and nanomanufacturing engineers, no longer recognizing divisions that
are not nature-made but man-made. It is this nanoengineering, done by chemists,
physicists, materials scientists, and engineers to put nanoscience into technology,
that is the theme of this book.
REFERENCES
1. Gibbs, J.W., Collected Works, Vol. 2, Longmans, New York, 1928, p. 262.
2. For example, Steven Chu, Claude Cohen-Tannoudji, and William D. Phillips, Physics;
1997; Ernst Ruska, Gerd Binnig, and Heinrich Rohrer, Physics 1986; Robert F. Curl
Jr., Sir Harold Kroto, and Richard E. Smalley, Chemistry; 1996.
3. Feynman, Richard P., Theres Plenty of Room at the Bottom, Talk at the annual
meeting of the American Physical Society at the California Institute of Technology,
December 29, 1959
4. Rowlinson, J.S., Faraday Lecture, The Molecular Theory of Small Systems, Chem-
ical Society Reviews, 12, 3, 1983; Rowlinson, J.S. and Widom, B., Molecular Theory
of Capillarity, Dover Publications, Mineola, NY, 1982 (2002 ed.).
5. Maguire, J., Talley, P., and Lupkowski, M., Journal of Adhesion, 45, 269290, 1994.
6. van der Waals, J.D., Z. Phys. Chem., 13, 657, 1894; English translation in J. Stat.
Phys., 20, 197, 1979; also Lord Rayleigh, Phil. Mag., 33, 209, 1892.
7. Lupkowski, M. and Maguire, J., Phys. Rev. B, 45, 23, 1992.
8. Maguire, J. and Leung, C.P., Phys. Rev. B, 43, 7, 1991.
9. Lupkowski, M. and Maguire, J., Composite Interfaces, 2, 1, 114, 1994.
10. Maguire, J., Benedict, M., LeClair, S., and Woodcock, L.V., Proc. Mater. Res. Soc.,
700, 241, 2002.
11. http://www.almaden.ibm.com/vis/stm/atomo.html;http://
www.almaden.ibm.com/vis/stm/corral.html
12. Hirano, K. et al., Appl. Phys. Lett., 80, 515, 2002.
13. For example, Myatt, C.J., Newbury, N.R., Ghrist, R.W., Loutzenhiser, S., and Wieman,
C.E., Opt. Lett., 21, 290, 1996.
14. For example, Ensher, J.R., Jin, D.S., Matthews, M.R., Wieman, C.E., and Cornell,
E.A., Phys. Rev. Lett., 77, 1996, 1996.
15. For example, McFarland, A.D. and Van Duyne, R.P., Nano Lett., 3, 10571062, 2003.
16. Sanchez, E.J., Novotny, I., and Xie, X.S., Phys. Rev. Lett., 82, 4014, 1999.
17. Pettinger, B., Rin, B., Picardi, G., Schuster, R., and Ertl, G., Phys. Rev. Lett., 92,
096101, 2004; D. Mast, Final Report, SBIR Phase I, Contract Number: #504624-
Sub Air Force, Advanced Adaptive Optical Coating Process Technologies, 01/04/02.
18. Saksena, R., Woodcock, L.V., and Maguire, J., Mol. Phys., 102, 3, 259266, 2004;
Maguire, J. and Benedict, M., Phys. Rev. B, 70, 174 112, 2004.
19. See A New Approach to Monte Carlo Simulation: Wang-Landau Sampling, D.P.
Landau, S.H. Tsai, and M. Exler, Amer. J. Phys.
20. de Gennes, P.G., Phys. Lett., 42, L-377, 1981; de Gennes 1991, in Nobel Lectures,
Physics, 19911995, Gsta Ekspong, Ed., World Scientific, Singapore, 1997.
21. Knight, T.A. and Woodcock, L.V., Test of the equipartition principle for granular
spheres in a saw tooth shaker, J. Phys. A Math & Gen., 29, 43654386, 1997.
22. http://www.nsf.gov/od/lpa/forum/colwell/rc81202.htm
23. Maguire, J.F., Contributions to Materials Science and Engineering, D.Sc. thesis,
University of Ulster, 2004; http://www.nsf.gov/od/lpa/forum/colwell/rc81202.htm.
2
Design
of Nanostructured
Materials
Debasish Banerjee, Jingyu Lao, and Zhifeng Ren
CONTENTS
2.1 Introduction
2.2 Motivation, Background, and Strategies
2.3 Experimental Setup
2.4 Results and Discussion
2.4.1 ZnO Nanostructures and Their Properties
2.4.2 The In
2
O
3
Nanocrystal Chain and Nanowire Circuit
2.4.3 Zn-In-O Hierarchical Nanostructures
2.4.4 Zn-Sn-O Hierarchical Nanostructures
2.4.5 Zn-In-Ge-O Quaternary Hierarchical Nanostructures
2.5 Large Quantity Nanostructures
2.6 Concluding Remarks
Problems
References
2.1 INTRODUCTION
Downsizing of existing microstructures to nanostructures (those structures in which
at least one dimension lies between 1 and 100 nanometers
1
(nm)) is found in widely
varied research fields. It possesses the potential to become one of the most fruitful
and farsighted scientific and technological innovations of recent decades. The
journey toward nanoscale science began in 1905, when Albert Einstein published
a paper estimating the diameter of a sugar molecule to be about 1 nm. In spite of
several important developments and a vague understanding of small-scale science,
this field remained dormant until critical technological progress was achieved.
Progress includes developments such as significant improvements in high-magni-
fication electron microscopy, the development of digital electronics, and greatly
improved chemical and physical control and manipulation capabilities. Although
development of the electron microscope, which enabled subnanometer imaging,
emerged around 1931, interest in nanotechnology was stimulated only later on the
occasion of Richard Feynmans
2
remarks in his 1959 landmark talk on aspects of
miniaturization entitled, Theres Plenty of Room at the Bottom. Feynmans
prediction of an age of nanoscience served as a lightning rod for innovative thinking
and set the pace for the evolution of todays nanotechnology. Among the early
milestones, the design in 1981 by Binnig and Rohrer of the tunneling scanning
microscope (STM), which can image an individual atom, provided further encour-
agement to researchers everywhere. This development led to accelerated studies in
small-scale science, stimulated further by the discovery of C
60
by Curl, Kroto, and
Smalley
3
and the discovery of the carbon nanotube by Iijima
4
in the mid-1980s and
early 1990s, respectively.
Contemporary scientists, engineers, futurists, and investors foresee nanoscale
science as a pathway to a vast field of smaller, faster, more energy-efficient tech-
nology through the expansion of their successes in downsizing microelectronics
and other cutting-edge developments. The unique properties
5
of miniaturization
have already demonstrated benefits in information storage,
6
in energy storage,
7
in chemical storage,
8
in chemical and biological sensing devices,
9
and over the
entire areas of electronics, photonics, and communication.
10
These technological
efforts have simultaneously enriched our understanding of physical phenomena
through direct observation of novel behavior in low-dimensional systems, includ-
ing size dependence, size-dependent photon emission (or photo-excitation),
11
coulomb blockage (or single electron tunneling),
12
metal-insulator transitions,
13
and quantized (or ballistic) conductance.
14
Quantum confinement
15
of electrons
in nanometer-sized structures may provide a powerful tool for controlling and
functionalizing the electric, optical, magnetic, and thermoelectric properties of these
advanced materials. Several unique nanostructures, including nanotubes,
16
nano-
wires, nanobelts,
17
and nanodots,
18
have been synthesized in laboratories worldwide.
They serve as building blocks for prototypes of many high-performance and ultra-
fast devices.
5,17,18
Here we provide a brief account of the strategies for growing various non-carbon-
based nanostructures and a description of our contributions to their synthesis and
characterization. We present the findings of recent measurements of some of the
physical and chemical properties
22,23
of complex hierarchical nanostructures,
1921
nanowalls,
22
nanowires,
21,23
and the self-assembly of nanocrystal chains and nanow-
ire circuits
24
in various systems. In a separate section, we consider a strategy for the
production of large quantities of these structures in freestanding versions to meet
the challenges of practical applications.
25,26
Lastly, we consider some possible future
applications.
2.2 MOTIVATION, BACKGROUND, AND STRATEGIES
Developments in contemporary microelectronics have reached a point where further
progress is limited by requirements for faster computing capabilities and by funda-
mental issues with the conventional top-down approach to fabrication, as well as with
its high cost. Significant advances toward improved production efficiency and precise
control over the morphology and microstructure of nanoscale materials through self-
assembly have been achieved. This opens a totally new, more cost-effective, more
economical, and more versatile bottom-up approach to fabrication. Furthermore,
self-assembly in nanoengineering has considerable potential for providing both high-
density fabrication capabilities and broad flexibility in choosing superior materials
for improved device functionality. Also, enhancement of the controllability and tuning
capacity of material properties can be accomplished by downsizing, powered by
quantum confinement effects.
30
In this area, several research groups have demon-
strated prototype basic electronic devices such as field effect transistors (FET), bipolar
junction transistors, p-n junctions, and resonant tunneling diodes
27
using nanowires
or other nanotube-based systems.
28
This approach becomes particularly powerful
when the ease and control offered by self-assembly is combined with the electronic,
magnetic, or photonic properties of inorganic components.
29
Accelerated progress in
nanotechnology beyond silicon will require substantial elevation of capability in
designing and manufacturing various complex three-dimensional (3-D) hierarchical
assembly nanostructures. We will comment further on our recent successes in syn-
thesizing novel 3-D hierarchical semiconducting nanostructures.
The growth of crystalline structures in low dimension
31
has been studied for
more than a century, but little of a quantitative nature has been accomplished.
Furthermore, the details of the control processes have not yet been mastered. The
general methodology is first, to concentrate the atomic or molecular building blocks
to form crystal nuclei that behave as seeds for growth, and second, to provide a
continuous and well-controlled environment in which the building blocks can assem-
ble the crystalline product. Control over the growth processes raises a nontrivial
issue. Perfection in high-quality crystals requires, in addition to a homogeneous
supply of building blocks, a reversible pathway between building blocks in the solid
phase surface and those in the liquid phase. These constraints provide for long-range
homogeneous enhancement of the building blocks ability to acquire the proper
orientation for forming a perfect crystal lattice with the selected morphology. Over
the past few years Lieber and coworkers, at Harvard University, and Yang et al., at
the University of California, Berkeley, have invented ingenious methods
3234
for
growing one-dimensional nanowires, rods, and belts. Materials with a highly aniso-
tropic crystal structure prefer one-dimensional growth because of the highly aniso-
tropic bonding between the building blocks in a chosen direction. Some polymeric
materials and biological systems, such as polysulfurnitride (SN)
x
, cellulose, and
collagen,
35
have high anisotropy and thus a preferred growth orientation. Other
anisotropic inorganic substances, including selenium, tellurium,
36
and the relatively
complex molybdenum chalcogenides,
37
which have a natural tendency to grow in a
preferred direction, have been studied extensively. Induced symmetry breaking is
required to achieve a preferred growth direction in isotropic crystals. Arc discharge,
laser ablation, and solution- and template-based
33
processes have been employed
for growing low-dimensional systems. The most general and convenient method,
however, has proven to be growth by a vapor phase transport process followed by
chemical vapor deposition.
34
Vapor phase epitaxy has thus become the most common
method for growing nanostructures of all solid materials. In principle any bulk
material can be converted to its one-dimensional nanostructure form through this
approach. In the early 1920s, Volmer and colleagues were the first to grow mercury
nanofibers by condensing mercury vapor on glass.
38
The fibers were 20 nm in diameter
and 1 Pm in length. Later, in the mid-1950s, a series of studies conducted by Sears
and colleagues determined that one-dimensional growth occurs through a general
mechanism based on axial screw dislocations.
39
This result appears to explain the
fundamental growth kinetics. We are now convinced that supersaturation during
growth is the key to acquiring diverse shapes and morphologies. The vapor-liquid-
solid (VLS) approach, first proposed by Wagner and Ellis in 1964, is the most widely
discussed process for growing one-dimensional nanowires. This procedure involves
the absorption of source material from the gas phase into a liquid droplet of a
catalyst
40
to form liquid alloy clusters. The liquid alloy clusters serve as preferential
sites for absorption and supersaturation of the reactant, thus breaking the symmetry.
This leads to the formation of one-dimensional nanowires. One-dimensional growth
occurs as long as the catalyst remains liquid. The most important issue with this
approach is to discover the perfect catalyst for a particular material, one that can
coexist as a liquid at the eutectic temperature. Observation of catalyst particles on
the ends of nanostructures is generally considered to be proof of VLS growth.
Semiconducting oxide nanowires of ZnO,
23,25,41
In
2
O
3
,
42
SiO
2
, Ga
2
O
3
,
43
and nanobelts
of ZnO, SnO
2
, In
2
O
3
, CdO, and PbO
44
have been synthesized in recent years using
the vapor phase epitaxy approach. Lee and coworkers have produced semiconducting
Si nanowires at relatively high yields using an oxide-assisted growth (OAG) mecha-
nism, an approach claimed to be fundamentally different from the VLS method.
45
In this oxide-assisted growth, the external metal catalyst is not necessary. Semiliquid
suboxide droplets, which have a lower melting point than the oxides, act as a catalyst
for the preferential growth sites.
We have already discussed the utility of ordering nanostructures in useful and
relatively complex forms. In the past few years, the nanoscience research com-
munity has achieved reasonable success in growing and applying one-dimensional
nanostructures, but few studies on hierarchical 3D structures have appeared. This
is most likely because of their complex mechanism and the difficulties in process
control. One simple approach for growing multidimensional nanostructures is first
to grow nanowires by the VLS method and then alter the synthesis conditions to
grow different materials on the surface of the nanowires by chemical vapor
deposition (CVD). Lauhon and coworkers demonstrated the synthesis of Si/Ge
and Ge/Si coaxial nanowires
46
and have shown that the outer shell can be formed
epitaxially on the outer core. Wang and colleagues
47
generated a mixture of coaxial
and biaxial SiC-SiO
x
nanowires via a catalyst-free high-temperature reaction of a
mixture of amorphous silica and graphite. Earlier, Zhu and coworkers
48
synthesized
3D flower-like nanostructures consisting of SiO
x
fibers radially attached to a single
catalytic particle by solidsolid or gassolid reaction. Using the same method,
they also synthesized a MgO fishbone assembly and a fernlike fractal nanostruc-
ture. Several other nontrivial 3D structures, such as Bi
2
S
3
skeletons
49
and SnO
50
nanoribbon networks, have also been demonstrated. Yang and colleagues showed
ZnO nanocrystals with exotic shapes, including tetrapods, tripods, and nanocombs,
synthesized by thermal evaporation of mixtures of ZnO and graphite.
51
In a similar
way, Wang and coworkers produced nanocombs and nanowindmills of wurtzite,
ZnS.
52
2.3 EXPERIMENTAL SET-UP
We have used a vapor phase epitaxyoriented growth process, which employs vapor
transport and condensation to grow complex microstructures. The process is rela-
tively simple and powerful for preparation of one-dimensional as well as more
complex 3-D structures, even in large quantities. First, we thermally evaporate from
either a mixed source or two different sources and then direct the vapor molecules
(building blocks for nanostructure growth) toward the substrate, which is at a lower
temperature. The vapor condenses on a suitable substrate, such as graphite foil or
sapphire, to form the nanostructures. The morphologies of the nanostructures are
controlled by varying the experimental conditions. We will describe the particular
changes separately in the corresponding sections.
The basic experimental set-up is illustrated schematically in Figure 2.1.
53
A
powder source is placed at the sealed end of an open quartz tube. The open end
of the quartz tube is covered with a graphite foil to serve as the substrate. The
entire assembly is inserted into a high-temperature horizontal ceramic tube fur-
nace in which a steep temperature gradient is established. To maintain the desired
internal atmospheric pressure between 1 militorr and 10 torr, one end of the
furnace tube is attached to a rotary vacuum pump while the other end is closed
by a silicon stopper connected to an adjustable needle valve. The temperature of
the furnace is ramped from ambient to a temperature between 900 to 1100qC at
a rate between 60 and 80qC/min. The quartz tube is positioned in the furnace so
that it is exposed to the appropriate temperature gradient. The substrate temper-
ature is also controlled by its location in the temperature gradient. Because of
the temperature gradient and the pumping direction, vapor produced from the
heated source condenses and deposits on the graphite foil. At the end of the
process, a fine powder is usually found deposited on the graphite foil surface
facing the vapor stream, as indicated by the coating visible in Figure 2.1. The
powder is analyzed using a JEOL JSM-6340F scanning electron microscope
(SEM), a Bruker Analytical X-ray System, and a JEOL 2010 transmission electron
microscope (TEM).
FIGURE 2.1 Experimental setup to grow nanostructures via vapor condensation and vapor
transport process.
Heating elements
Quartz tube
Substrate
Ceramic tube
Mixed powder
source
2.4 RESULTS AND DISCUSSION
Simple thermal evaporation and condensation of various mixtures of semiconductor
oxides can effectively synthesize very complex and useful nanostructures. The precise
growth mechanism of these nanocrystal formations is still obscure because of the
complex transport characteristics of various building blocks and the intricacy of
orientation-related hetero and homoepitaxial crystallographic matching among dif-
ferent materials. Consequently, control over morphology and size is far from satis-
factory. In the sections that follow, we will concentrate on the interesting binary
semiconducting materials zinc oxide (ZnO), indium oxide (In
2
O
3
), and tin oxide
(SnO
2
). ZnO is a direct wide band gap (3.37 eV) semiconducting material with
significantly higher exciton binding energy (60 meV) than that of its competitors,
GaN (25 meV) and ZnSe (22 meV). This fact deserves substantial attention because,
to achieve efficient excitonic lasing action at room temperature, the excitonic binding
energy must be higher than the thermal binding energy at room temperature (26 meV).
ZnO naturally is a promising candidate to be a room-temperature short-wavelength
laser.
10
Low-dimensional systems, such as nanostructures, have lower threshold lasing
energy as a consequence of quantum effects that create a substantial density of states
near the band edges. Carrier confinement enhances radiative recombination.
54
Thus
ZnO nanostructures have a significant potential for opto-electronic applications. Apart
from electronic and opto-electronic
19,41(a)
applications, ZnO has demonstrated several
other optical and electro-optical applications in optical waveguides,
55
optical
switches,
56
transparent ultraviolet protective conducting films,
57
and acousto-optic
and surfaceacoustic applications.
58
ZnO also is a well-established sensor material
59
that has demonstrated utility in solar cells
58
and photo-catalysts.
61
Among the other
wide band gap transparent materials, the interesting optical and electronic properties
of In
2
O
3
have drawn considerable attention. In
2
O
3
is also a direct wide band gap (3.6
eV) material with various applications in solar cells, organic lightemitting diodes,
and gas sensors.
62
2.4.1 ZnO NANOSTRUCTURES AND THEIR PROPERTIES
ZnO nanostructures can easily be synthesized with predetermined size and morphol-
ogy through the application of vapor phase epitaxy. Chemical vapor deposition
(CVD),
63
metal organic chemical vapor deposition (MOCVD),
64
physical vapor
deposition (PVD),
65
and thermal evaporation
25,26,66
represent alternative methods for
growing ZnO nanowires on various substrates. All of these have been explored in
recent years. Control over orientation and size of the nanowires has also been
achieved.
26,41(a)
A popular method for producing ZnO nanowires through CVD or
PVD is to evaporate either pure Zn, a mixture of Zn and another material such as
Se, or a mixture of ZnO and graphite in the proper proportion as a source material
in a furnace similar to that shown in Figure 2.1. In this synthesis the vapor would
be directed toward the substrate in a flow of argon or a mixture of oxygen, argon,
and hydrogen. In the MOCVD technique, metal organic zinc precursors such as
diethyl zinc (Zn(C
2
H
5
)
2
) or zinc acetylacetonate hydrate (Zn(C
5
H
7
O
2
)
2
u H
2
O) are
normally used. ZnO nanowires can also be produced through the simple thermal
evaporation of a mixture of ZnO and graphite while maintaining an appropriate oxygen
partial pressure in the tube furnace and condensing the vapor on a selected substrate
utilizing the temperature gradient along the tube. Alignment and size control of the
nanowires can be achieved by employing an appropriate custom-designed substrate.
Nanowires of certain materials have a preferred direction of growth. For example,
Si nanowires prefer to grow along the 110 direction, whereas ZnO nanowires prefer
the 001 direction. Therefore the proper strategy for growing vertically aligned
nanowires is to select the appropriate thin film substrate with a suitable material to
serve as a catalyst. When the substrate is heated to the melting temperature of the
catalyst, it melts. Because of the high surface free energy, small particles appear
that condense and form clusters of alloy. These supersaturate with the incoming
vapor and become a one-dimensional nanostructure through the VLS mechanism as
discussed previously. The diameter of the nanowires can be adjusted by controlling
the thickness of the film. The preferred substrate for Si and ZnO nanowires is
either a silicon or sapphire substrate with a thin Au coating as the catalyst. One
may use an Si (111) wafer as a substrate for the growth of vertically aligned Si
nanowire arrays,
45
whereas a-plane (110) sapphire can serve to grow epitaxial vertical
ZnO nanowires.
41(a)
For a more comprehensive discussion of these procedures, please
browse the respective references.
Epitaxial growth of vertically aligned ZnO nanowires on an a-plane sapphire
substrate with Au coating is shown in Figure 2.2. In the initial stage of the formation
of nanowires, the basis for the formation of aligned arrays of nanorods, zinc-gold alloy
deposits in separated dots on the substrate. The SEM image of a large array of the
aligned ZnO nanorods is shown in Figure 2.2(a), and Figure 2.2(b) is a top view at
medium magnification. In this case, only the heads of the nanorods are visible. The
nanorods are 50 to 120 nm in diameter, with a length between several hundred
FIGURE 2.2 Aligned ZnO nanowires grown on a-plane sapphire substrate. Top view of
aligned hexagonal ZnO nanowires at low (a) and medium (b) magnifications, and tilted view
(c).
nanometers and a few microns. Figure 2.2(c) is a tilted view of the aligned ZnO
nanorods. For field emission and other electronic applications, the use of a sapphire
substrate is not convenient because of its low conductivity. Crystalline Si substrate
is preferable instead. Gold nanoparticles, 3 nm in diameter, are spread on the silicon
substrate to serve as a catalyst. This procedure also presents a problem, because Au
nanoparticles are very mobile at high temperatures, and aggregation may occur. A
summary of the results we have obtained is shown in Figure 2.3. In Figure 2.3(a) a
ZnO nanowire has grown on the Si surface. The nanowires usually have a diameter
between 10 and 20 nm and a length of several microns. A high-magnification view
of such nanowires is illustrated in Figure 2.3(b). Some larger nanowires are also
visible. The formation of these large nanowires is caused by the aggregation of Au
nanoparticles. Figure 2.3(c) shows the large aggregated Au nanoparticles and ZnO
nanowires nucleated from these nanoparticles. Gold nanoparticles have even aggre-
gated on the side of the nanowire.
ZnO nanowalls have been synthesized under similar conditions on 1- to 3-nm-
thick Au coated a-plane sapphire substrate. In this case the substrate temperature
was held between 875 and 950qC, with the tube pressure between 0.5 and 1.5 torr.
Nanowalls grown at high temperature show a milky white-gray color, whereas
structures grown at low temperature show a reddish color. The SEM images of the
morphology of typical ZnO nanowalls are shown in Figures 2.4(a), through (c). The
nanowalls are interconnected to form a network, with most of the nanowall flakes
normal to the substrate. The pore size varies from 100 nm to 1 Pm. These nanowall
flakes do not exhibit a clearly ordered pattern but rather show a quasi-hexagonal
pattern with most of the flakes forming angles that are multiples of 30q. The x-ray
FIGURE 2.3 Formation of ZnO nanowires on an Si substrate coated with few nanometers
of gold (Au). Low (a) and high (b) magnification SEM image of the small ZnO nanowires
formed on the Si substrate where gold droplets acts as catalysts, and shows aggregation
problem of Au clusters at high processing temperaturec(c).
2theta diffraction patterns of the nanowalls are shown in Figure 2.5(a). As a result
of the good epitaxial relation between the c-plane of the ZnO nanowalls and the
a-plane of the sapphire, only the ZnO (0002) and (0004) peaks can be seen. In the
omega scan the peak splits with a full width at half maximum (FWHM) of 0.03q as
displayed in Figure 2.5(b). This split could originate from the Si substrate itself. A
similar split (with similar FWHM) in the bare Si substrate can be observed (Figure
2.5(c)). The phi scan of the (0112) peak of a ZnO nanowall on sapphire is shown
FIGURE 2.4 SEM images of the ZnO nanowalls synthesized by vapor transport and conden-
sation method. (a) A medium magnification SEM image of the small size nanowalls. (b) A
medium magnification SEM image of the large size nanowalls. (c) A high magnification SEM
image of the large size nanowalls.
in Figure 2.5(d). The six peaks, with equivalent distances of 60q demonstrate the
in-plane epitaxial relation of the ZnO nanowalls with the a-plane single crystal
sapphire substrate. The TEM images of nanowall flakes reveal that the atomic
orientations are somewhat inconsistent with the presence of a significant number of
dislocations. There is a low-magnification view of such a nanowall flake in Figure
2.6(a), and, in Figure 2.6(b), an associated electron diffraction pattern is shown. The
appearance of extinction diffraction spots such as (0001), (0003), and so forth, is
caused by double diffraction. From the SAD pattern, the nanowall flake is on the
(1110) plane, and the direction from bottom to top of the plot is [0001], consistent
with the XRD examination (Figure 2.5(a)). Dislocations on the surface of the
flake can easily be identified by the parallel lines shown in Figure 2.6(c). These
FIGURE 2.5 XRD spectra of the nanowalls structure. (a) Theta2 theta scan. (b) and (c) Omega
scan of the nanowalls and substrate, respectively. (d) Phi scan of the nanowalls.
dislocations, with a length of about 20 nm, distribute periodically along the flake.
The high-resolution image shown in Figure 2.6(d) indicates that each dislocation
line is associated with two heavily strained areas. If one draws a Burgers circuit to
enclose each heavily strained area, one can see that each strained area is associated
with a perfect dislocation and a Burgers vector of either 1/3(1 0) or 1/3( 110).
The fact that each dislocation line is associated with two dislocations with opposite
Burgers vectors suggests that these dislocations are actually dislocation dipoles.
The Burgers vectors of these unit-perfect dislocations are parallel to the (0002) plane,
typical mobile dislocations for a hexagonal structure. The reason for the existence
of dislocations in the flakes is not yet clear.
FIGURE 2.6 TEM micrographs of the structure of the nanowalls. (a) A low magnification
TEM image of a nanowall flake. (b) An SAD pattern. (c) A high magnification phase contrast
image showing the edge dislocation dipoles. (d) High-resolution TEM image of the dislocation
dipole.
11 2
Photoluminescence (PL) studies of ZnO nanowalls were conducted at room
temperature using a dual scanning microplate spectrofluorometer (Molecular
Devices Inc., SpectraMax Gemini XS) with an excitation wavelength of 325 nm.
Emission scanning was conducted over the range between 340 and 600 nm. The PL
spectra of ZnO nanowalls is shown in Figure 2.7. The white-gray nanowalls (curve
a) grown at high temperature show significantly different PL spectra from those of
the reddish nanowalls (curve b) grown at a lower temperature. Nanowalls grown at
higher temperature exhibit a strong UV peak (380 nm), attributed to the band-edge
emission of ZnO wurtzite, with a weak deep band green-yellow peak. Reddish
samples grown at lower temperature have a weak UV peak with a large, broad,
green-yellow peak. The green-yellow is attributed to defects (e.g., oxygen vacancies)
that result in a radiative transition between shallow donors and deep acceptors.
67
It
is widely accepted
68
that the green peak of ZnO is due to radiative recombination
of photo-generated holes and electrons associated with oxygen vacancies. Annealing
in an oxygen atmosphere at 600qC for 2 h suppresses the green-yellow emission
completely, and vacuum annealing at 400qC suppresses both the UV and green-
yellow emission by a small amount. The strong emission from the band edge suggests
that the nanowalls grown at higher temperature have an excellent crystalline struc-
ture, consistent with the other reports on nanowires.
41(a)
Because of the high aspect ratio in nanotubes and nanowires, a thin film of such
nanostructures is found to be an excellent field emission electron source. Among other
nanostructures, nanotube thin films have already been studied in great detail.
69
With
better chemical stability and structural rigidity, ZnO nanowires are expected to provide
a more stable and better field emission electron source than do nanotubes. For inves-
tigation of the field emission properties of ZnO nanowires, we used two separate
methods for preparing the Au catalysts on a 100 substrate. One method was thermal
evaporation of a 1 to 3 nm Au film, and the other was dispersion of 3 nm Au
nanoparticles in toluene on the Si substrate. The production of Au nanoparticles
FIGURE 2.7 Photoluminescence spectra of the nanowalls. (a) Spectrum of white-gray
nanowalls grown at high temperature and (b) spectrum of reddish nanowalls grown at low
temperature.
employed tetraoctylammonium bromide, as a transfer agent to move Au ions from
an aqueous environment to toluene, and dodecylamine, as a capping molecule for
the crystalline nanoparticles.
71
The suspension was diluted to different concentrations
in toluene for different areal densities of Au nanoparticles on the Si surface. Eight
samples with different Au catalyst densities, as described in Figure 2.8(e), were
prepared in this study. Two of these samples were prepared using a continuous Au
catalyst thin film, and the remaining six samples were prepared with four different
concentrations of the Au catalyst nanoparticle suspension. A mixture of ZnO and
FIGURE 2.8 SEM micrographs of (a) sample B, with continuous Au film, (b) sample D,
with density 5.6 u 10
6
/cm
2
, (c) sample F with density 1.2 u 10
6
/cm
2
, and (d) sample H, with
density 0.4 u 10
6
/cm
2
. (e) Table showing different areal density of various substrates. (f) The
measured current densities as a function of the macroscopic electric field for 8 samples. The
measurements were carried out at a vacuum level of about 2 u 10
6
torr.
graphite was used as a source material. Some of our SEM images of these samples
are shown in Figure 2.8. A ZnO nanowire thin film formed from the continuous Au
film deposited on the silicon substrate is shown in Figure 2.8(a). Figures 2.8(b),
2.8(c), and 2.8(d) show the corresponding ZnO nanowire thin films generated from
the Au nanoparticles with areal densities on the Si substrate of 5.6 u 10
6
, 1.2 u 10
6
,
and 0.4 u 10
6
/cm
2
, respectively. The field emission current of all these samples was
measured using a simple diode configuration.
71
A plot of the measured current
densities as a function of the microscopic electric field is shown in Figure 2.8(f ).
The horizontal line corresponds to a current density of 1 mA/cm
2
, and the values of
electric field required to obtain this current density are 18.77, 18.50, 15.57, 14.96,
12.92, 11.43, 10.16, and 6.46 V/Pm for samples A through H, respectively. The
pressure was 2 u 10
6
torr during the measurement, much lower than is usually
employed.
72
Therefore, it is expected that a much lower macroscopic electric field
could have been achieved at a somewhat higher pressure if the same screening effect
observed on carbon nanotube field emitters also affects the field emission from thin
films of ZnO nanowires. Thin films with the lowest areal density of ZnO nanowires
showed much stronger field emission characteristics, comparable to those of carbon
nanotubes. The field emission characteristics of ZnO nanowire thin films were
further improved on annealing in hydrogen.
Under experimental conditions similar to those in the nanowire synthesis, we
found some nanocomb-like structures at a relatively higher temperature, as shown
in Figure 2.9. This is consistent with other previous reports
51
of comb-like ZnO
nanostructures that appear when pure Zn is used as the evaporation source. The
large amount of such structures is seen in Figure 2.9(a) with a high-magnification
image in Figure 2.9(b). Interestingly, in Figure 2.9(b), the nanostructure grows
along the [1120] direction, which is different from the usual [0110] direction
common in the hierarchical structure (discussed in Section 2.4.3) formed from
binary sources. We explored some other types of nanostructures, where nanocrystals
were attached to nanowires (see Figure 2.10). The small ZnO nanowires of tens of
nanometers in diameter are decorated with nanocrystals (Figures 2.10(a) and
2.10(b)). Some of the nanocrystals are attached epitaxially to the nanowire. An
unusual structure appears where a nanorod meets the nanobelt during growth and
penetrates through it (Figure 2.10(c)).
FIGURE 2.9 ZnO comb-like nanostructures. (a) Low magnification and (b) high magnifica-
tion SEM images.
2.4.2 THE In
2
O
3
NANOCRYSTAL CHAIN AND NANOWIRE CIRCUIT
Transition from a top-down to a bottom-up approach in nanoelectronics through
self assembly, as we have mentioned, requires a significant research effort and a
profound understanding of the influences of external parameters on the growth of
nanostructures. When air was introduced into the tube furnace, we found that In
2
O
3
nanostructures grown by vapor transport and condensation are greatly influenced
by the atmospheric pressure in the growth zone and by the indium or indium
suboxide vapor pressure. By controlling the internal atmospheric pressure, it is
possible to substantially change the morphology from a nanowire structure to a
nanocrystal or a nanowire chain in circular form. We used a mixture of In
2
O
3
and
graphite placed near the sealed end of the quartz tube as the source material.
Native SiO
2
coated Si (SiO
2
/Si), on which a 1 to 3 nm Au thin film was deposited,
formed on the substrate, which was kept close to the open end of the quartz tube
and loosely sealed. Finally, the entire assembly was introduced into a quartz tube
furnace, pumped, and heated to a temperature between 1000 and 1030qC, and held
for 15 to 30 min. The pressure inside the tube during growth was controlled by a
gauge valve. When the substrate temperature is between 875 and 950qC and the
pressure is between 0.1 and 1.0 torr, In
2
O
3
nanostructures grow on the SiO
2
/Si
substrate.
In
2
O
3
nanowires can be grown in this setup at a pressure of 1.0 torr maintained
from the beginning of heating. Scanning electron microscopy (SEM) images of
the In
2
O
3
nanowires on the SiO
2
/Si substrate usually obtained from the lower
temperature end of the substrate are shown in Figure 2.11(a). The In
2
O
3
nanowires,
with a typical body-centered cubic structure, have diameters ranging from 15 to
FIGURE 2.10 Figure 2.6 (a) and (b) ZnO nanocrystal decorated nanorods and nanowires.
(c) A ZnO nanorod penetrated nanobelt.
60 nm and lengths of tens of micrometers. A TEM image and a high-resolution
transmission electron microgram (HRTEM) image of a nanowire having a [001]
growth direction are seen in Figures 2.11(b) and 2.11(c). The Au particle that
serves as a catalyst is clearly visible on the tip of the nanowire. We observed an
increase of pressure from 0.05 torr to a peak of about 0.13 torr with continuous
pumping in a tight system (no air inlet). This occurred because of the generation
of indium vapor followed by a decline due to exhaustion of the In
2
O
3
source. If
air is introduced at the time when the Indium vapor pressure has already declined
from a peak of 0.13 to about 0.06 torr, we obtained various complex self-assembled
crystal chains and circuit networks (Figure 2.12). In this case we introduced an
air pressure of 1.0 at this low indium vapor pressure (about 0.06 torr). The
nanocrystal chains have diameters 20 to 100 nm and lengths of a few micrometers
(Figure 2.12(a)).
FIGURE 2.11 SEM and TEM microscopic images of the In
2
O
3
nanowires. (a) Medium
magnification SEM image of the nanowires. (b) TEM image of the nanowire. (c) HRTEM
image showing the nanowire with an Au catalyst on the tip.
The nanocrystal chains are connected together in a complex circuit network.
The network junctions can be either nanocrystals or larger crystals. Based on our
knowledge of crystal growth and our observation of the detailed morphologies, we
find the reasonable growth direction of this particular nanocrystal chain must be
1-2-3-4-5, as shown in the figure. On the way from point 4 to point 5, two additional
FIGURE 2.12 SEM and TEM microscopic images of the In
2
O
3
nanowire and nanocrystal
chain circuits. Big crystals are part of the circuit. (a) SEM image showing the nanocrystal
chain circuits. (b) SEM image showing the circuit junctions. (c) SEM image showing the
nanowire and nanocrystal circuits. (d) TEM bright field image of part of a nanocrystal chain.
(e) SAD pattern corresponding to the nanocrystal on the left of point X and (f) on the right
of point X. (g) HRTEM showing the domain boundary at point X.
nanochains were bounced off of it. The large crystals are purely cubic and have a
solid connection (see Figure 2.12(b)) with another crystal. Hence they are available
to be used as leads for a nanoelectronic circuit. Figure 2.12(c) illustrates some
networks where there are solid nanowires instead of crystal chains. The TEM image
of part of a nanocrystal chain is visible in Figure 2.12(d). The chain is formed by
the interconnection of an individual nanocrystal with a growth direction of [001]
with the nanocrystals epitaxial to one another. The structures of these nanocrystals
are perfect, and no amorphous layer is found on the surface. The chain can change
geometric direction when it meets with an obstacle. It is found that the change of
the chains geometric direction, however, is caused by a change in crystal orientation,
not by meeting an obstacle. The turning point, X, is indicated by an arrow in Figure
2.12(d). Electron diffraction patterns from the right and left of point X are shown
in Figures 2.12(e) and 2.12(f ), respectively. The chain crystal growth direction is
always along the [001] direction. Figure 2.12(g) is an HRTEM image taken from
point X. The domain boundary and change of lattice orientation can be clearly seen
between two nanocrystals.
The nanowire circuits where few large crystals exist after a 0.3 torr pressure of
air was introduced into the low indium vapor pressure of 0.06 torr are seen in a
SEM image. Figures 2.13(a) and 2.13(b) show the SEM image of such circuits.
These nanowires, having a diameter of 20 to 100 nm and a length of tens of
micrometers, grow in a zigzag fashion. After running into another nanowire, the
initial nanowire changed direction and continued growth. Figure 2.13(c) shows an
interesting hexagonally shaped nanowire circuit with other nanowires growing inside
FIGURE 2.13 SEM images of the In
2
O
3
nanowire circuits grown at 0.3 torr pressure. (a) Low
magnification SEM image showing the circuit. (b) Medium SEM image showing the junctions.
(c) A hexagonally shaped circuit. (d) SEM image showing the parallel nanowires from a big
nanofiber. Some nanowires with zigzag growth direction can be seen. (e) High magnification
SEM image showing the zigzag growth direction of a nanowire confined between the two
parallel nanowires.
the hexagon. Many small nanowires are seen growing along the side of an In
2
O
3
large fiber (Figure 2.13(d)). These nanowires are parallel to each other at the start
and change growth direction after a certain length. Small particles are visible at the
tip of many nanowires. A small zigzag nanowire confined between two nanowires
is illustrated in Figure 2.13(e).
Under the same experimental conditions, but without the Au catalyst layer,
many In
2
O
3
microcrystals were formed on the substrates. A few interesting In
2
O
3
nanocrystal chains were also found on some areas with an In/In
2
O
3
film deposited
on the SiO
2
/Si substrate. Figures 2.14(a) and 2.14(b) show some wavy nanocrystal
chains on the In/In
2
O
3
film. All the nanocrystal chains are nucleated from the edge
of the holes in the film. One In
2
O
3
nanocrystal chain is seen growing directly out
of the hole (Figure 2.14(b)). Another chain changed direction after meeting with
the hole edge and then grew out of the hole and ran into another hole. It then
changed direction after meeting with the edge of the smaller hole and finally grew
out of the hole. One nanocrystal chain changed growth direction three times at the
hole edge before growing out of the hole (Figure 2.14(c)). A chain is seen forming
into a circle confined to the edge of a smaller hole before growing out (Figure 2.14(d)).
The nanostructures grown using the vapor-liquid-solid (VLS) mechanism
73
usu-
ally have either a constant diameter or a periodically changing diameter
74
along the
growth direction, whereas here we see nanostructures composed of many nanocrys-
tals with different sizes. Therefore, it is likely that the nanocrystal chains shown in
Figure 2.12 are created through a vapor-solid (VS) mechanism. The higher atmo-
spheric pressure over the indium oxide leads to these crystal chains, which is also
consistent with the VLS mechanism. It is important to emphasize that a gold catalyst
is necessary for the formation of these circuits. Formation of nanowires instead of
FIGURE 2.14 SEM images of the wavy In
2
O
3
nanowires grown without Au catalyst. (a) Low
magnification SEM image. All the nanowires are started from the edge of the holes. (b) Wavy
nanowire crossed between two holes. (c) A wavy nanowire changed direction three times in
a hole. (d) A wavy nanowire ring formed along the edge of a hole.
nanochains is probably a consequence of the temperature gradient in the furnace. It
is fascinating to find that the force behind the nanowire/nanocrystal chain formation
is so strong that, in most cases, the nanowire continues to grow albeit with a direction
change after meeting obstacles. It is generally expected in consideration of the VS
mechanism that the deposited molecules on the side of the nanowire will move to
the growth front and contribute to the one-dimensional nanostructure growth.
Because of the limited diffusion length of the molecules deposited on the side of
the nanocrystals, not every nanocrystal will exhibit direct materials exchange. These
nanocircuits have immense potential as building blocks for nanoelectronic devices.
2.4.3 Zn-In-O HIERARCHICAL NANOSTRUCTURES
Hierarchical nanostructures with 6-, 4-, and 2-fold symmetries can be grown using
the experimental set-up shown in Fig. 2.1 as described in Section 2.3. ZnO nanorods
can be grown on an In
2
O
3
nanowire core with various symmetries and orientations
by vapor transport and condensation. ZnO and In
2
O
3
have high melting points, (i.e.,
1,975qC and 1,931qC respectively). We therefore mixed in graphite powder as a
reducing agent, which brings down the reaction temperature to below 1,000qC. A
mixture of ZnO, In
2
O
3
, and graphite powders can be used as a coevaporation source
to be placed at the sealed end of a single-ended sealed quartz tube (shown in Fig.
2.1). The nanostructures can grow on many different collectors, including graphite
foil, single crystal silicon, and LaAlO
3
. In our experiment, mixed powders were
heated between 950 and 1000qC for 30 min. in the ceramic tube under a pressure
between approximately 0.5 and 2.5 torr. The vapor was transported after evaporation
to the open end of the inner quartz tube, where it was oxidized in the presence of
air and condensed on the graphite foil. (Other substrates, such as silicon and LaAlO
3,
also have been used.) The oxidation rate was controlled by the amount of air leaked
into the tube furnace through a needle valve connected to the open end of the ceramic
tube. A region of relatively sharp temperature gradient was chosen so that the
condensation temperature could be controlled between 820 and 870qC. The flux
ratio of In alloying agent to Zn in the vapor phase was controlled by mixing the
desired amount of In
2
O
3
with ZnO powder in the original source. It has been
determined that the In:Zn flux ratio is the key to control of the preferred symmetry
in the hierarchical nanostructure. For example, a higher In:Zn flux ratio tends to
grow six 4-fold symmetric nanostructures, whereas medium and lower In:Zn flux
ratios grow 2-fold symmetric or nano ribbon structures.
The SEM images of the ZnO nanostructures with In
2
O
3
as the alloying agent
at low and medium magnifications are illustrated in Fig. 2.15. Large quantity
hierarchical ZnO nanostructures are shown in Fig. 2.15(a). Under medium mag-
nification, Fig. 2.15(b) clearly shows all the major 6-, 4-, and 2-fold structural
symmetries. The length of the major In
2
O
3
core is usually on the order of tens of
microns, and the diameter is between 50 and 500 nm. The secondary ZnO nanorods
on the core surfaces are about 0.2 Pm in length and possess a diameter between
20 and 200 nm. They are oriented either normal to or at an angle to the core
surface. This path branches out to a few more subsymmetries associated with the
FIGURE 2.15 SEM images of the ZnO nanostructures synthesized by vapor transport and
condensation technique. (a) Low magnification SEM image of the ZnO nanostructures to
show the abundance. (b) Medium magnification SEM image of ZnO nanostructures to show
the various structural symmetries. Three major basic symmetries of 6-, 4-, and 2-fold were
clearly seen. (c) XRD observation of the ZnO hierarchical nanostructures.
major symmetries. Three subsymmetries, namely 6S-, 6M-, and 6M
*
, have been
observed for 6-fold major symmetries. Similar subsymmetries can also be found
with 4-fold (4S-, 4S
*1
-, 4S
*2
-, 4M-, and 4M
*
) and 2-fold (2S-, 2S
*
, and 2M-)
symmetries respectively. Powder x-ray diffraction (XRD) reveals that the sample
is a mixture of hexagonal wurtzite ZnO and cubic In
2
O
3
, as shown in Figure
2.15(c). The lattice constants for ZnO have been calculated to be a 3.249 and
c 5.206 , and for cubic In
2
O
3
, a 10.118 . In both cases these values are in
agreement with the values for their bulk forms.
76
Detailed SEM and TEM obser-
vations of 6-fold nanostructures associated with various subsymmetries are found
in Figure 2.16. Figures 2.16(a) and 2.16(c) show, at medium and high magnification,
the majority of 6S-fold nanostructures possessing smaller core diameters. As the
nomenclature suggests, they have a single row of secondary ZnO nanorods on the core
surface. If the available surface area is large enough, they can grow in multiple rows
with 6M-fold symmetry as shown in Figures 2.16(b) and 2.16(d), in medium and
high magnification, respectively. It is reasonable to conclude that the hexagonal
symmetric surface of the major In
2
O
3
nanowire core is responsible for the exotic
secondary growth of ZnO nanorods. The hexagonal surface of the core is illustrated
in Figures 2.16(e) and 2.16(f ). Clearly it can be assumed, depending on the
available surface area on the each segment of the hexagonal core, that single or
multiple rows may grow. The formation of a 6M
*
-fold structure, where the rows
of secondary ZnO nanorods are at an angle to the In
2
O
3
major core, is visible in
Figure 2.16(g). An energy dispersive x-ray spectroscopy (EDS) study of the sample
demonstrates that the major core axis is pure In
2
O
3
, whereas secondary nanorods
are pure ZnO. This is further confirmed by the clearly evident contrast in the TEM
image in Figure 2.16(i). Orientation relationships between major core axes (In
2
O
3
)
and secondary nanorods (ZnO) can be studied by selected area diffraction patterns
(SAD). In each diffraction pattern, two separate sets of features are prominent, where
one is from the core and the other, is from the secondary rods. The diffraction pattern
in Figure 2.16( j) has been indexed using the [110] zone axis of In
2
O
3
and the [1120]
zone axis of ZnO. Hence the crystallographic relationship comes out to be ,
[1120]
ZnO
for the 6-fold symmetry. When In
2
O
3
is along the [110] direction, the core
nanowire is enclosed by r[112], r[112], and r[110] facets. The angle between each
of these adjacent facets is very close to 60q, so a quasi 6-fold symmetry is often
observed when an In
2
O
3
nanowire grows along the [110] direction. In addition to the
[001] and [110] directions, In
2
O
3
nanowires also grow along the [111] direction, as
shown in Figures 2.16(e) and 2.16(f ) (and illustrated in Figure 2.16(h)), where
hexagon end planes are clearly observable at the end of major cores and on secondary
ZnO. A detailed conclusion on the orientation relationship of 6-fold symmetry in
contrast to 4-fold symmetry is given later.
The various aspects of 4-fold symmetric nanostructures are found in Figure 2.17.
The majority formation is in 4-fold symmetry, as seen in Figure 2.17(a). As already
discussed, there are at least five possible varieties of 4-fold structures: 4S, 4S
*1
-,
4S
*2
-, 4M-, and 4M
*
. A closer look at 4S symmetry, where a single row of
secondary nanorods are normal to the core, is given in Figure 2.17(b), whereas in
*
Definitions of the symbols are found in Reference 75.
[110]
In O
2 3
FIGURE 2.16 SEM and TEM images, selected area diffraction patterns, and the schematic
growth models of the 6-fold ZnO nanostructures. (a) SEM image showing the abundance of
the 6S-fold symmetry. (b) SEM image showing the 6M-fold symmetry. (c) High magnification
SEM image of the 6S-fold symmetry. (d) High magnification SEM image of the 6M-fold
symmetry. (e) Head-on view of a 6S-fold symmetry to show the hexagonal nature of the major
core nanowire. (f ) Side view of the structure in (e) to show the hexagonal nature of all the
secondary ZnO nanorods and their same growth orientations with the major In
2
O
3
core
nanowire. (g) 6M
*
-fold symmetry, where the nanorods are not perpendicular to the major
core. (h) Schematic diagram of orientation relationships between the major In
2
O
3
core nanow-
ire and the secondary ZnO nanorods with the core along [110] direction (h left) and along
[111] direction (h) right). (i) Cross-sectional bright-field TEM image of 6-fold symmetry
showing the six facets of the central core. ( j) Selected-area electron diffraction pattern of ( i)
corresponding to the major In
2
O
3
core and the secondary ZnO nanorods.
Figure 2.17(c), multiple rows of perpendicular nanorods (4M) can be seen for the
same reason discussed for its 6-fold counterpart. Similar to what was seen in the
6-fold case, in 4-fold nanostructures, secondary ZnO nanorods are not always
perpendicular to the major In
2
O
3
core nanowire but rather grow at discrete angles
to the In
2
O
3
core nanowire. In Figure 2.17(e), a single row of secondary ZnO
FIGURE 2.17 SEM and TEM images, selected area diffraction patterns, and the schematic
growth models of the 4-fold ZnO nanostructures. (a) Medium magnification SEM image show-
ing the abundance of the 4-fold nanostructures. Scale bar, 5 Pm. (b) High magnification SEM
image showing the 4S-fold symmetry. (c) High magnification SEM image showing the 4M-fold
symmetry. (d) High magnification SEM image of the 4S
*1
-fold symmetry. (e) High magnification
SEM image of the 4S
*2
-fold symmetry. (f ) High magnification SEM image of the 4M
*
-fold
symmetry. (g) Schematic model of the 4S- and 4S
*
-fold symmetry. (h) Cross-sectional bright-
field TEM image of 4-fold symmetry to show the four facets of the central core. (i) Selected-
area electron diffraction pattern of (h) corresponding to the major In
2
O
3
core and the secondary
ZnO nanorods. The diffraction consists of two sets of patterns; the small rectangle corresponds
to [001] zone axis of In
2
O
3
while the large solid one corresponds to [6392] zone axis of ZnO.
The dashed rectangle is from another arm perpendicular to the solid rectangle.
nanorods grow at the same angle in all four directions. In Figure 2.17(d), the
growth is seen to exist only in the two opposing directions (parallel to the page).
The other two opposing directions are perpendicular to the major In
2
O
3
core
nanowire (into and out of the page). We define these two nanostructures as 4S
*1
-
and 4S
*2
-fold, respectively. When the major In
2
O
3
core nanowire is large enough,
we can even see multiple rows of ZnO nanorods growing at an angle in all four
directions (4M
*
-fold symmetry) as shown in Figure 2.17(f ). Figures 2.17(h) and
2.17(i) show TEM bright field images and selected area diffraction patterns, which
reveal, in contrast to 6-fold symmetry (with hexagonal core), that the major core
is of cubic In
2
O
3
nanowires, and the secondary rods are hexagonal ZnO nanorods
for the 4-fold case. In a bright field TEM image, Figure 2.17(f ) shows four facets
of the central 4-fold symmetric In
2
O
3
core. The diffraction patterns in Figure
2.17(g) consist of two sets. The small rectangle corresponds to [001] zone axis of
In
2
O
3
, whereas the large solid one corresponds to the [6392] zone axis of ZnO.
The dashed rectangle is from a different arm perpendicular to the solid rectangle.
Figure 2.17(g) is the schematic model of 4S- and 4S
*
-fold growth. The orientation
relationship in the basic 4-fold symmetry is as follows: //[0110]
ZnO
,
//[1120]
ZnO
. When an In
2
O
3
nanowire is along the [001] direction, the
core nanowire is enclosed by r (100) and r (010) facets. In this case, all ZnO
nanorods in the four arms grow perpendicular to the core nanowire because ZnO
nanorods grow along the [0001] direction. They can be expressed symbolically
as A [0001]
ZnO
. In the case of the tilted secondary nanorod with 4-fold
symmetry, the angle between a ZnO nanorod and a core nanowire (around 60q) is
equal to the angle between [6392]
ZnO
, and [0001]
ZnO
. Because there is no difference
between and [6392]
ZnO
for the growth of a nanorod on a [001] core, In
2
O
3
nanowire nanorods can grow at an angle of either 60 or 120q. This results in all
the variations of tilted growth. The heteroepitaxial nature of ZnO nanorods from
In
2
O
3
cores provides many possible crystal orientation relations among the cores
and nanorods. This is the source of so many different ZnO nanorod orientations with
respect to the core. Thus the symmetry of these hierarchical nanostructures is depen-
dent on the crystallographic orientation of the In
2
O
3
core nanowires. The orientation
of an In
2
O
3
nanowire along the [110] or [111] direction creates all the 6-fold sym-
metries, whereas along the [001] direction, all the 4-fold symmetries are produced.
The orientation relationships demonstrated previously in Figures 2.16(h) and 2.17(g)
are listed in Table 2.1 and can be understood through the theory of coincidence site
lattices. For example, the In
2
O
3
a-plane has 4-fold symmetry with a 10.18 . The
ZnO c-plane has 6-fold symmetry with a 3.24 . This results in a lattice mismatch
of about 3.7% (a factor of 3 for the In
2
O
3
a-axis to the ZnO a-axis), a reasonable
value for epitaxial growth.
In addition to 4-fold and 6-fold nanostructures, 2-fold symmetry has also been
observed under similar experimental conditions. SEM images of 2-fold symmetric
nanostructures appear in the In/Zn system. It was mentioned earlier that a medium
In : Zn flux ratio in the source material enhances the growth of the 2-fold symmetric
structures. Typical 2S-fold symmetric structures, where single perpendicular row
nanorods grow on the two sides of the core, are shown in Figures 2.18(a) and
2.18(b). As previously stated, multiple rows may also be grown depending upon the
[ ] 001
In O
2 3
[ ] 001
In O
2 3
[ ] 001
In O
2 3
availability of sufficient surface area on the core. Figures 2.18(c) and 2.18(d)
illustrate typical 2M-fold symmetric structures. 2S
*
-fold structures are also possible,
where secondary nanorods make an angle with the core, as demonstrated in Figures
2.18(e) and 2.18(f ). If the In : Zn flux ratio is further reduced, wool-like nanorib-
bons, with lengths ranging from 10 to 20 Pm and widths from 50 to 200 nm, have
been observed. A HRTEM study [18(b)] reveals that these In-alloyed ZnO nanor-
ibbons consist of a central nanowire core with two adjacent nanoribbons flanking
it, in contrast to the previously observed [15(c)] pure ZnO nanoribbon specimen.
The TEM image of a single ribbon, where the central core and the two adjacent
ribbons can easily be distinguished, is seen in Figure 2.18(h). With increasing
In : Zn flux ratio, additional growth of nanorods perpendicular to the plane of the
ribbons (in the [0001] direction) can be found. As in Figure 2.18(i), the SEM image
indicates that these nanowires grow only from the central core of the nanoribbons
and then self-align to form a straight row of nanorods. An EDS study indicates that
the nanorods and nanoribbons grew epitaxially on the nanowire core.
We have already suggested that the variation of In : Zn flux ratio is crucial in
controlling the morphology of the Zn-In-O hierarchical nanostructures. By reducing
the In
2
O
3
content in the source, we found a few more exotic nanostructures in the
Zn-In-O system, including nanobridges, nanonails, and nanopins. The experimental
conditions were the same except that we set the atmospheric pressure in the furnace
to 2.0 torr for nanobridges. For nanonails it was held between 0.5 and 1.0 torr. The
temperature of the furnace was 1000qC, and the condensation temperature was
approximately between 950 and 970qC. A summary of the SEM and TEM obser-
vations of nanobridges defined as geometrical analogs to ancient stone bridge struc-
tures is shown in Figure 2.19. A typical nanobridge can be tens of microns long and
up to a few microns in width and height (Figure 2.19(a)). The rows of nanorods
grow at the edge of both sides of the nanobelts with varying density and are usually
directed perpendicular to the belt surface (Figure 2.19(b)). It is even possible to
TABLE 2.1
Observed Orientation Relationships between the Major
In
2
O
3
Core Nanowire and the Secondary ZnO Nanorods
Core Axis Orientation Relationship
(Figure 2.16(h)) //[112 0]
ZnO
//[0001]
ZnO
(Figure 2.16(h)) //[1010]
ZnO
//[1210]
ZnO
A [0001]
ZnO
(Figure 2.17(g)) //[1010]
ZnO
//[1210]
ZnO
A [0001]
ZnO
(Figure 2.17(g)) //[6392]
ZnO
//[1210]
ZnO
//[1013]
ZnO
[ ] 110
In O
2 3
[ ] 110
In O
2 3
[ ] 222
In O
2 3
[ ] 111
In O
2 3
[ ] 111
In O
2 3
[ ] 112
In O
2 3
[ ] 110
In O
2 3
[ ] 001
In O
2 3
[ ] 001
In O
2 3
[ ] 100
In O
2 3
[ ] 001
In O
2 3
[ ] 001
In O
2 3
[ ] 001
In O
2 3
[ ] 110
In O
2 3
[ ] 110
In O
2 3
grow multiple rows of nanorods. The nanorods are 50 to 200 nm wide and 2 Pm
long. Figure 2.19(c) is a side view of the TEM image of a nanobridge, with the
selected area diffraction pattern (SAD) illustrated in the inset. The nanobelt has a
diffraction pattern similar to that of the nanorods except that its high-index diffraction
spots are split. It is well established that the individual hexagonal nanorod grows
FIGURE 2.18 SEM and TEM images of the 2-fold nanostructures. (a) Medium SEM image
to show the abundance of the 2S-fold symmetry. (b) High magnification of 2S-fold symmetry.
(c) Low magnification SEM image to show the abundance of the 2M-fold symmetry. (d) High
magnification SEM image of 2M-fold symmetry. (e) Low magnification SEM image to show the
abundance of the 2S
*
-fold symmetry. ( f ) High magnification SEM image of 2S
*
-fold symmetry.
(h) SEM image of In-alloyed ZnO nanoribbons grown at low In: Zn ratio (i). TEM image of a
nanoribbon with showing dark contrast between the core and the side ribbons. ( j ) SEM image
of a nanowire core with both nanoribbons and nanorods, where nanorods are well aligned and
normal to the growth of side nanoribbons.
epitaxially along the [0001] direction on the (0001) plane and/ or along the [0001]
direction on the (0001) plane of the belt. The nanorod rows are parallel to the
[0110] direction of the belt. The zone axis of the diffraction pattern in the insert
is [2110], as illustrated in Figure 2.19(d). Explanations of the various symmetries
and of the versatility of the nanobridge growth we have found thus far are given in
Figure 2.20. In Figure 2.20(a) the belt is shown to form a central ring, with the
nanorods growing along the belt edges exactly perpendicular to the belt surface.
This behavior is due to the strong epitaxial relation between the nanobelt and
nanorods. Figure 2.20(b) shows a rare situation where two belts join together per-
pendicular to each other, with the associated nanorods growing according to their
original direction perpendicular to their own belt. The availability of an edge surface
on the belt determines the nanorods growth habit. For example, in Figure 2.20(c)
nanorods grew on all four sides of the belt. In total, eight rows of nanorods can be
observed with 4-fold symmetry. Figure 2.20(d) shows another variation. The density
of the nanorods (perpendicular to the plot) on the belts varies considerably. One
edge has high density, whereas the other edge has low density. In addition, there is
FIGURE 2.19 SEM images of the ZnO nanobridges synthesized by vapor transport and
condensation method. (a) Low magnification image showing the abundance of the nanobridges.
(b) Medium magnification image of top view of a nanobridge. (c) TEM image of the side view
of a nanobridge. The insert is the electron diffraction pattern of a nanorod, with zone axis of
[2110] direction. Scale bar = 200 nm. (d) Schematic drawing of top and side view of part of a
nanobridge. The page planes for the top view and side view are (0001) and (1120), respectively.
another row of nanorods growing parallel to the belt surface at an angle of about
36q to the belt growth direction.
We found additional interesting structures, such as nanonails and nanopins, by
slightly varying the experimental conditions. The SEM and TEM studies of such
nanonails are described in Figure 2.21. A low-magnification view of the aligned
nanonails, together with some nanonail flowers, can be found in Figure 2.21(a). The
inserted XRD pattern shows the strong (002) peak of the wurtzite ZnO structure
and much weaker (100) and (101) peaks due to the imperfect vertical growth of
the nanonails. Figure 2.21(b) shows a side view of the nanoflowers to illustrate
their hemispheric shape. Figures 2.21(c) and 2.21(d) are medium- and high-
magnification views of the aligned nanonails. These nanonails are several microns
in length, with a shaft diameter of 150 to 200 nm. They form roughly perpendicular
to the surface. The hexagonal cap is approximately 1 Pm in diameter and 50 to
100 nm thick. The change of diameter from shaft to cap is abrupt. The density of
nanonails is high. They form quasi-perpendicular to the substrate, a habit that is
probably a consequence of the epitaxial growth of ZnO along the c-axis of the
graphite foil substrate. A TEM side view displays an image of a nanonail that
exhibits dark contrast because it is too thick to be transparent to the electron beam
(Figure 2.21(e)). The inserted SAD pattern confirms that the nanonail grows along
the c-axis. No splitting of the diffraction spots has been found here. A high-
resolution TEM (HRTEM) image taken from the cap of the nanonail demonstrates
the perfect lattice structure of the nanonail and also confirms that the nanonail
possesses wurtzite structure (Figure 2.21(f )). The EDS spectrum shows that the
nanonails are composed only of zinc and oxygen, which means that no indium
contamination has been found. Similar nanonails of a much larger size were made
previously
66
by vaporizing a mixture of only ZnO and graphite powder in a similar
FIGURE 2.20 SEM images of ZnO nanobridge variations. (a) Roller coaster-like and (b)
joined twin nanobridges; (c) and (d) are both combinations of a nanobridge and 4-fold
symmetry.
experimental set-up. In a different region we found very small (100 nm cap
diameter with 1 Pm length) nanonails (or nanopins). The SEM images of such
nanonails in low and high magnification are illustrated in Figures 2.22(a) and
2.22(b). The diameter decreases to 40 nm at the end. Figures 2.22(c) and 2.22(d)
illustrate SEM images of another type of nanonail where the size reduction from
bottom to top is much larger. In this case the diameter is continuously reduced
from 800 nm at the top to 20 nm at the bottom with a perfect hexagonal cap at
the top. In yet another set of nanonails, we find that the caps are not perfectly
FIGURE 2.21 (a) Low-magnification SEM image of the ZnO nanonails synthesized by vapor
transport and condensation method showing the aligned growth of nanonails and the nanonail
flowers. Insert is the x-ray diffraction pattern. (b) Medium-magnification view of nanonail
flower. Scale bar = 5 Pm. (c) Side view showing the vertical growth of nanonails. (d) High-
magnification side view SEM image of a nanonail. (e) TEM image of the nanonail. The insert
is the electron diffraction pattern, with zone axis of [2110]. (f ) HRTEM image taken from
the cap of the nanonail along [2110] direction.
hexagonal, although they have hexagonal symmetry as indicated by the six facets
(Figure 2.22(e)). These nanowires generally grow on ZnO nanorod bases (Figure
2.22(f )). There is also evidence that nanonails can grow on a thin ZnO sheet
(Figure 2.22(g)).
The growth process of these Zn-In-O hierarchical nanostructures can be
extremely complex because of their binary transport mechanism and complicated
crystallographic orientations. The reduction of oxides, vapor transport, vapor oxi-
dation, and condensation at the low-temperature collector are also involved.
Because we have not used any catalyst in this process, the growth mechanism
cannot be described as being of VLS origin, but rather it may follow a vapor-solid
route.
77
In the ZnO-In
2
O
3
-graphite system, first both ZnO and In
2
O
3
are probably
reduced to suboxides (ZnO
x
or InO
x
[x 1]), which have much lower melting points.
FIGURE 2.22 SEM images of several different ZnO nanonail structures. (a) Low and
(b) medium magnification images of small nanonails. (c) Medium and (d) high magni-
fication images of thin shaft nanonails. (e) Non-hexagon shape nanonails on ZnO rod
bases. ( f ) Nanonails on ZnO sheet.
Between these two suboxides, InO
x
vaporizes first and then, because of the tem-
perature gradient along the furnace, condenses on the collector, where it deposits
hexagonal, tetragonal, or rectangular In
2
O
3
nanowires. ZnO vaporizes later and then
condenses on the existing In
2
O
3
nanowire core to form a layer of ZnO with 6-, 4-
, and 2-fold symmetries. These provide the basis for further growth of ZnO nanorods
through the vapor-solid mechanism. Cubic In
2
O
3
prefers to grow along the [001]
direction to form in 4-fold symmetry, while it chooses the [110] or [111] direction
for 6-fold symmetry. Because we observed nanorods prematurely deposited on the
In
2
O
3
core surface, we conclude that it is very unlikely that both the core and
secondary structures grow simultaneously. Because premature growth of nanorods
on the In
2
O
3
core surface in the early stage of nucleation occurs, we believe random
fluctuations in the early stages of nucleation might initiate growth. Formation of
ZnO nanoribbons with a central core has been studied in detail as previously
reported.
20
A TEM study reveals the existence in the ribbons of a (ZnO)
11
In
2
O
3
nanowire core. The presence of this alloying element is necessary for the growth
of side branching, which suggests that, during growth of the central nanowire, the
alloying element continuously migrates to the facets of the nanowire core. The
alloy-rich layers of the facets become nucleation sites for subsequent growth of the
side branches by forming 2-fold symmetric structures. Zinc suboxide and liquid
zinc metal, both of which have much lower melting points than ZnO, dominate the
formation of nanobridges and nanonails. From the EDS and HRTEM studies of the
nanobridges, it is seen that nanorods are composed of pure ZnO, and the belt is
ZnO dominated, contaminated with only 0 to 3% atomic indium. This suggests that
the growth of a ZnO nanorod from the nanobelt is homoepitaxial. We believe
formation of the Zn-In-O eutectic phase significantly contributes to the nanobridge
structure, because it has been observed that without In
2
O
3
, the evaporation rate
becomes very low. The reason that nanorods grow on the edges is somewhat
unclear, but a high surface free energy probably plays a role. For nanonails, ZnO
x
and liquid Zn metal dominate. Following nucleation, small nanorods sprout epi-
taxially from the graphite foil substrate or from ZnO
x
structures. The incoming
ZnO
x
vapor deposits epitaxially on both the shaft and the cap of the nanorod sprout.
The nanorod cap, having a higher surface area, has the better chance to absorb
ZnO
x
vapor than does the bottom region. The absorbed ZnO
x
vapor increases as
the diameter of the hexagonal cap increases. Further, ZnO
x
vapor is absorbed at
the bottom. This causes the entire structure to be elevated. The diameter often
continuously decreases from top to bottom. This is not always the case, however,
because other factors, such as vapor dynamics and the oxidation rate, occasionally
play a critical role.
2.4.4 Zn-Sn-O HIERARCHICAL NANOSTRUCTURES
As in the Zn-In-O system, hierarchical nanostructures can also be observed in the
Zn-Sn-O system. A mixture of ZnO, SnO
2
, and graphite powder has been used as
a source for synthesizing Sn doped ZnO nanostructures. The experimental condi-
tions and set-up are similar to those previously described (see Figure 2.23 for SEM
images of these structures). The majority of the product is nanobelt-like, as shown
in Figure 2.23(a). On closer examination we find that it consists of either straight
or twisted nanorods welded together, as shown in Figures 2.23(a) and 2.23(b). Small
nanorods occasionally grow from the middle core onto the surface of the nanobelts,
which are also the junctions of the elemental nanorods. The elemental nanorods
can either appear perpendicular to the nanobelt growth direction as shown in Figure
2.23(d) or form at other angles. Figure 2.23(c) shows a nanobelt with three rows
of nanorods growing on one side of the nanobelt surface. The middle row of the
nanorods is perpendicular to the nanobelt surface, and the other two rows appear
to form at 45. We speculate that such a structure would possess 8-fold symmetry
if secondary nanorods also grow on the other side of the nanobelt surface. Unlike
the Zn-In-O nanobridge structure, no growth of secondary nanorods has been
observed on the edges of the nanobelts. Two-fold symmetric structures, which may
be the premature stage of the nanobelt formations, are shown in Figure 2.23(e). In
Figure 2.23(e), only one side of the core nanorods has secondary nanorods, whereas
there appears in Figure 2.23(f ) a very rare tree leaf-like structure. It is likely that
the Zn-Sn-O nanostructures are formed in two steps. The Zn-Sn-O core nanobelt
forms first, and then the secondary nanorods grow epitaxially on the nanobelt under
FIGURE 2.23 SEM image of the hierarchical Zn-Sn-O nanostructures. (a) Low-magnifica-
tion and (b) high-magnification image of the nanobelts. (c) An 8-fold structure. (d) A 4-fold
structure. (e) A 2-fold structure. (f) A rare leaf-like structure.
the right conditions. Most likely all the structures are based on homoepitaxials due
to the dominant number of bare nanobelts and the similarity of the core nanobelt
to hierarchical nanostructures with major bare nanobelts. Based on these observa-
tions, we believe there are several other missing hierarchical structures to be found
in the Zn-Sn-O system.
2.4.5 Zn-In-Ge-O QUATERNARY HIERARCHICAL NANOSTRUCTURES
Success with binary and ternary systems encouraged us to explore some quaternary
systems such as Zn-In-Ge-O. A mixture of ZnO, GeO
2
, In
2
O
3
, and graphite powder
was used as a source. Initial experiments produced some complicated ordered
hierarchical nanostructures as shown in Figures 2.24(a) and 2.24(b). The secondary
nanorods grew disordered on the major core nanorod. A higher magnification image
of the structure is shown in Figure 2.24(b). The structures usually are tens of microns
in length, and the core nanorods are in the range of 100 nm in diameter. The
secondary nanorods are smaller in diameter, with the length varying between 200
nm and 2 Pm.
2.5 LARGE QUANTITY NANOSTRUCTURES
From the preceding sections of this discussion, it is quite apparent that significant
progress has been achieved in producing several exotic and useful nanostructures
for various systems. It is important for practical purposes, as well, to have the ability
to supply large quantities in a freestanding form. Most of the nanostructures we have
discussed were grown on a substrate such as graphite foil, single crystal silicon, or
sapphire. Growth of nanostructures on any two-dimensional substrate, however, does
not yield large quantities. Rather, one must devise a process that will produce
freestanding nanostructures in gram quantities at least. In this section we shall
describe our recent success in producing such large quantity, freestanding ZnO
nanowires. The method we have employed can easily be extended to any of the other
nanostructures discussed in preceding sections.
The set-up we used is similar to that shown in Figure 2.25 except that the quartz
tube was replaced by a quartz boat. The entire boat was loosely covered by a quartz
FIGURE 2.24 Microscopy studies of Zn-Ge-In-O hierarchical nanostructure. (a) Low-mag-
nification and (b) high-magnification SEM images.
plate with a nozzle hole at the front to introduce air flow. One end of the furnace
was connected to a rotary pump to regulate the air pressure between 0.1 and 10
torr, according to the specific requirements of the process. The other end was closed
by a silicone stopper connected to an adjustable needle valve. A mixture of ZnO
and graphite powder was placed at one end of the boat close to the nozzle hole,
while between 2 and 3 cm from the source material, micron-size graphite flakes
were spread uniformly along the direction of air flow to serve as a medium for
collecting the nanowires. The boat was positioned within the steep temperature
gradient of the furnace. At the end of the process, a substantial quantity of nano-
wires were found to have grown on the high surface area of the graphite flakes.
The yield was substantial. The weight of the graphite flakes was more than twice
their original weight, and 40% of the original ZnO was converted to nanowires. A
SEM view of the product is shown in Figure 2.26. Nanowires as long as 15 Pm
formed in the relatively high temperature zone of the furnace near the source,
whereas at a lower temperature, far from the source, the nanowires were only 2 to
5 Pm long, with diameters between 70 and 100 nm. Figures 2.26(a), (b), (c), and
(d) show such long nanowires under various magnifications. They possess smooth
round surfaces with very sharp tips. The shorter nanowires have blunt hexagonal
heads, with a hexagonal surface (Figures 2.26(e), (f ), (g), and (h)). Under XRD
analysis, these nanowires were proven to be entirely of the wurtzite hexagonal
structure, similar to that of bulk ZnO. Their unit cell constants are as follows: a
3.248 and c 5.206 . These results were later confirmed by selected area
diffraction (SAD). Further, an HRTEM study confirms that the nanowires are perfect
crystals without dislocations, slaking, or other defects. The nanowires were separated
from the graphite flakes to acquire freestanding nanowires by oxidizing the sample
in flowing O
2
at 700qC. The XRD study shown in Figure 2.27 confirms that total
removal of the graphite content was achieved by this procedure. More convincing
proof of removal is seen in the SEM image of the oxidized sample, which shows
voids that mimic the shape of the flakes (Figure 2.28(c)). Under TEM observation,
the surface of the nanowires is nearly always wrapped in amorphous carbon or
turbostratic graphite (Figure 2.28(a)) which was completely removed through the
oxidation process as can be seen in Figure 2.28(b). Occasionally the oxidation
FIGURE 2.25 Experimental set-up to grow large quantity nanostructures via vapor transport
and condensation method.
Heating elements
Quartz boat
Quartz plate
Ceramic tube
Mixed powder
source
process causes the nanowires to become less electrically conductive. Fortunately,
the conductivity can be restored by further annealing at 500qC in a vacuum.
The high surface area of the graphite flakes is the key to large quantity growth
of ZnO nanowires. Other nanopowders, such as Al
2
O
3
, have not yielded similar
quantities. Success in producing large quantity ZnO nanowires raises the prospect
of synthesizing the other previously observed nanostructures in large quantities as
well. Our initial experiments in that direction have already begun with interesting
positive results. A detailed study of the variation of yield with different growth
FIGURE 2.26 Morphology of as-made ZnO nanowires shown using SEM images. (a), (b),
(c) and (d) show long (1015 Pm) and thin (3060 nm) nanowires grown at the higher
temperature region of the furnace whereas (e), (f ), (g), and (h) show short (12 Pm) and fat
(60100 nm) nanowires grown at the lower temperature region of the furnace. The scale bar for
(a) is 10 Pm, for (b), (c), (e), (f ), and (g) is for 1 Pm and for (d), and (h) is 100 nm.
FIGURE 2.27 XRD pattern recorded from ZnO nanowires grown via vapor transport and
condensation method.
FIGURE 2.28 Reduction of amorphous graphite layer through annealing process. (a) TEM
image of the surface of an as-made nanowire with amorphous graphite shell. (b) TEM
image of an oxidized ZnO nanowire with no amorphous graphite shell on it. (c) SEM image
of the oxidized ZnO nanowires, voids mimicking the shape of the graphite flakes. Scales
bars are (a) 2 nm, (b) 5 nm, (c) 1 Pm.
2-Theta (degrees)
15 20 30 40 50 60 70
0
100
200
300
400
500
600
700
800
900
R
e
l
a
t
i
v
e

i
n
t
e
n
s
i
t
y

(
a
.
u
.
)
A
s
-
m
a
d
e
O
x
i
d
a
t
i
o
n

a
t

6
0
0

d
e
g
r
e
e

C
O
x
i
d
a
t
i
o
n

a
t

6
5
0

d
e
g
r
e
e

C
O
x
i
d
a
t
i
o
n

a
t

7
0
0

d
e
g
r
e
e

C
G
r
a
p
h
i
t
e

(
0
0
2
)
G
r
a
p
h
i
t
e

(
0
0
4
)
G
r
a
p
h
i
t
e

(
1
0
0
)
G
r
a
p
h
i
t
e

(
1
0
1
)
(
1
0
0
)
(
0
0
2
)
(
1
0
1
)
(
1
0
2
)
(
1
0
3
)
(
2
0
0
)
(
1
1
2
)
(
2
0
1
)
(
1
1
0
)
(
0
0
4
)
parameters, such as pressure and temperature in the tube furnace, the amount of
graphite in the original mixture of the source, and the source-to-collector ratio, has
been conducted and will be reported shortly.
26
2.6 CONCLUDING REMARKS
We have discussed the synthesis and analysis of nanostructures (nanowires, nano-
walls, nanocircuits, nanochains, and others) composed of binary systems, notably
ZnO and In
2
O
3
; ternary systems, especially Zn-In-O and Zn-Sn-O; and significant
modifications of both, including exotic ternary hierarchical systems and quaternary
Zn-In-Ge-O hierarchical structures. We have described their morphology and struc-
ture as well as their electrical and optical properties. At this point our detailed
understanding of their chemical and physical fundamentals is far from complete.
Complex epitaxial relationships, with numerous possible orientations and relations
among different components of these mixed systems, complicate access to a straight-
forward understanding. We do believe, however, that we are making progress. Our
results make us certain that the vapor transport and condensation process is a powerful
tool for the production of a large variety of nanoscale self-assemblies aimed at meeting
the challenges of a bottom-up approach to nanoengineering. We also described our
initiative growing these nanostructures in freestanding gram quantities. We find that
the important matter of gaining precise control over size and morphology for practical
applications requires a significant effort toward a more profound, detailed understand-
ing. All the results we have discussed are very recent. This inspires us, in pursuit of
direct practical applications, to continue close observation of our results as we go
forward. We are certain that these nanostructures have great potential in a variety of
fields including field emission; photovoltaics; transparent EMI shielding; super capac-
itors; fuel cells; high-strength, multifunctional composites; and many other branches
of nanoelectronics and optoeletronics. Last, we must recognize that this is a truly
interdisciplinary field encompassing physics, chemistry, materials science, and engi-
neering. Our immediate goals will be to fully master the synthesis of the assemblies
with reliably predictable morphology and physical properties.
PROBLEMS
*
1. What are the key advantages of the bottom-up approach to top-down
approaches in nanoscale fabrication?
2. Discuss the strategies one can adopt to grow nanostructures out of their
bulk form.
3. Discuss the vapor phase epitaxy approach for nanostructure growth.
4. How does the vapor-solid-liquid (VLS) growth mechanism of nanostruc-
tures work? What are its advantages? What is the key difference between
VLS and OAG (oxide-assisted growth)?
5. What are the strategies for growing nanostructures in complex hierarchical
form? Explain and give examples.
*
Readers are also advised to read corresponding references to answer the questions.
6. Why is ZnO superior to other oxide semiconductors for shortwave lasing
applications? What are the advantages of ZnO in its nanostructure over
its bulk form?
7. Discuss applications of ZnO and In
2
O
3
in various fields.
8. What are the various techniques of producing nanostructures? Explain the
vapor transport and condensation technique. What are the useful charac-
terization techniques for nanostructures?
REFERENCES
1. H. Gleiter, Nanostructure materials: Basic concepts and microstructures, Acta Mater.,
48, 1, 2000.
2. R.P. Feynman, Miniaturization, Reinhold, New York, 1961.
3. H.W. Kroto, J.R. Health, S.C. OBrian, R.F. Curl, and R.E. Smalley, Nature, 318,
6042, 1985.
4. S. Iijima, Nature, 354, 56, 1991.
5. a) A.S. Edelstein, and R.C. Cammarata, Eds., Nanomaterials: Synthesis, Properties,
and Application, Institute of Physics, Philadelphia, PA, 1996;
b) M. Grundmann, Ed., Nano-optoelectronics, Springer-Verlag, 2002; Miture, S.,
Nanomaterials, Pergamon Press, New York, 2000;
c) Lojkowsky, W., Ed., Interfacial Effects and Novel Properties of Nanomaterials,
Scitec Publications, 2003;
d) H.S. Nalwa, Ed., Hand-book of Nanostructured Materials and Nanotechnology,
Academic Press, New York, 2000.
6. a) H. Chopra, and S. Hua, Phys. Rev., B 66, 020403, 2002; b) Y.C. Kong, D.P. Yu,
B. Zhang, W. Fang, and S.Q. Feng, Appl. Phys. Lett., 78, 4, 2001; c) R. O Barr, S.Y.
Yamamoto, S. Schultz, W.H. Xu, and A. Scherer, J. Appl. Phys., 81, 4730, 1997.
7. a) R.H. Baughman, A.A. Zakhidov, W.A. de Heer, Science, 297 (5582), 787792,
2002; b) S.A. Chesnokov, V.A. Nalimova, A.G. Rinzler, R.E. Smalley, J.E. Fischer,
Phys. Rev. Lett., 82 (2): 343346, 1999; c) P.A. Gordon and P.B. Saeger, Industrial
& Engineering Chemistry Research, 38 (12): 46474655, 1999; d) K.H. An, W.S.
Kim, Y.S. Park, Y.C. Choi, S.M. Lee, D.C. Chung, D.J. Bae, S.C. Lim, Y.H. Lee,
Adv. Mater., 13 (7), 497, 2001.
8. a) S.M. Lee and Y.H. Lee, Appl. Phys. Lett., 76 (20): 28772879, 2000; b) R.H.
Baughman, A.A. Zakhidov, and W.A. de Heer, Science, 297 (5582): 787792, 2002.
9. a) F. Favier, E.C. Walter, M.P. Zach, T. Benter, and R.M. Penner, Science, 293 (5538):
22272231, 2001; b) Y. Cui, Q. Wei, H. Park, and C.M. Lieber, Science, 293, 1298,
2001.
10. a) R.F. Service, Science 1997, 276, 895; b) R.H. Baughman, A.A. Zakhidov, W.A.
de Heer, Science, 297 (5582): 787792, 2002.
11. A.P. Alivisator, Science, 1996, 271, 1993.
12. a) C. Schonenberger, H. Vanhouten, H.C. Donfersloot, Euro. Phys. Lett., 20 (3):
249254, 1992; b) Y. Tokura, D.G. Austing, S. Tarucha, J. Phys-Cond Matter, 11
(31): 60236034, 1999; c) N. Asahi, M. Akazawa, Y. Amemiya, IEEE Transaction
on Electron Devices, 44 (7): 11091116, 1997; d) D. Ali, H. Ahmed, Appl. Phys.
Lett., 64 (16): 21192120, 1994.
13. a) Z. Zhang, X. Sun, M.S. Dresselhaus, J.Y. Ying, Phys. Rev. B, 61, 4850, 2000;
b) G. Markovich, C.P. Collier, S.E. Henrrichs, F. Remacle, R.D. Levine, J.R. Heath,
Acc. Chem. Res., 32, 415, 1999.
14. J.M. Krans, J.M. van Rutenbeek, V.V. Fisun, I.K. Yanson, L.J. de Jongh, Nature, 375,
767, 1995.
15. D.J. Lockwood, Z.H. Lu, J.M. Baribeau, Phys. Rev. Lett., 76 (3): 539541, 1996.
16. a) S. Iijima, Nature, 354, 56, 1991; b) T.W. Ebbesen, P.M. Ajayan, Nature, 358, 220,
1992; c) Z.F. Ren, Z.P. Huang, J.W. Xu, J.H. Wang, P. Bush, M.P. Siegal, and P.N.
Provencio, Science, 282, 11051107, 1998.
17. a) M.S. Dresselhaus, Y.M. Lin, O. Robin, M.R. Black, G. Dresselhaus, Nanowires,
2003 (available the Mildred S. Dresselhauss Group Web site: http://eecs-pc-
05.mit.edu/x986.pdf; b) Z.L. Wang, Ed., Nanowires and Nanobelts: Materials, Prop-
erties and Devices, Kluwer Academic Publishers, 2003; c) Z.W. Pan, Z.R. Dai, Z.L.
Wang, Science, 291, 19471949, 2001.
18. S. Bandyopadhyay, H.S. Nalwa, Eds., Quantum Dots and Nanowires, American
Scientific Publication, 2003.
19. J.Y. Lao, J.G. Wen, Z.F. Ren, Nano Lett., 2(11), 1287, 2002.
20. a) J.Y. Lao, J.Y. Huang, D.Z. Wang and Z.F. Ren, Nano Lett., 3(2), 235, 2002; b)
J.G. Wen, J.Y. Lao, D.Z. Wang, T.M. Kyaw, Y.L. Foo, Z.F. Ren, Chem. Phys. Lett.,
372, 717722, 2003.
21. J.Y. Lao, J.Y. Huang, D. Banerjee, S.H. Jo, D.Z. Wang, J.G. Wen, D. Steeves, B. Kimball,
W. Porter, R.A. Farrer, T. Baldacchini, J.T. Fourkas, Z.F. Ren. SPIE Proceedings, 5219,
99, 2003; J.Y. Lao, J.Y. Huang, D.Z. Wang, J.G. Wen, Z.F. Ren, Novel Hierarchical
Oxide Nanostructures, J. Chem. Mater., 14, 770, 2004.
22. J.Y. Lao, J.Y. Huang, D.Z. Wang, Z.F. Ren, D. Steeves, B. Kimball, W. Porter, Appl.
Phys. A, 78, 539, 2004.
23. S.H. Jo, J.Y. Lao, and Z.F. Ren, R.A. Farrer, T. Baldacchini, J.T. Fourkas, Appl. Phys.
Lett., 82, 3520, 2003.
24. J.Y. Lao, J.Y. Haung, D.Z. Wang, Z.F. Ren, Adv. Mater., 16, 65, 2004.
25. D. Banerjee, J.Y. Lao, D.Z. Wang, J.Y. Huang, Z.F. Ren, D. Steeves, B. Kimball, and
M. Sennett, Appl. Phys. Lett., 83, 2062, 2003.
26. D. Banerjee, J.Y. Lao, D.Z. Wang, J.Y. Huang, D. Steeves, B. Kimball, and Z.F. Ren,
Nanotechnology, 15, 404, 2004.
27. a) X.F. Daun, Y. Huang, Y. Cui, J.F. Wang, and C.M. Lieber, Nature, 409, 66, 2001;
b) Y. Huang, X. Daun, Y. Cui, L.J. Lauhon, K.H. Kim, C.M. Lieber, Science, 294,
1313, 2001; c) D.H. Cobden, Nature, 409, 32, 2001; d) G.Y. Tseng, J.C. Ellenbogen,
Science, 2001, 294, 1293, 2001.
28. R.H. Baughman, A.A. Zakhidov, W.A. de Heer, Science, 297 (5582): 787792, 2002.
29. a) C.T. Black, C.B. Murray, R.L. Sandstrom, S. Sun, Science, 290, 11311134, 2000;
b) T. Thurn-Albrecht, Science, 290, 21262129 (2000); c) C. Sanchez, B. Lebeau,
Mater. Res. Soc. Bull., 26, 377387, 2001.
30. a) B.B. Li, D.P. Yu, S.L. Zhang, Phys. Rev. B, 59 (3), 16451648, 1999; b) S.Q. Feng,
D.P. Yu, H.Z. Zhang, Z.G. Bai, Y. Ding, J. Crys. Growth, 209 (23): 513517, 2000;
c) A.M. Rao, E. Richter, S. Bandow, B. Chase, P.C. Eklund, K.A. Williams, S. Fang,
K.R. Subbaswamy, M. Menon, M.A. Thess, R.E. Smalley, G. Dresselhaus, M.S.
Dresselhaus, Science, 275 (5297): 187191, 1997.
31. a) R.J. Davey and J. Garside, From Molecules to Crystallizers, Oxford Chemistry
Primers, No. 86, Oxford University Press, Oxford, 2001; b) P. Knauth, J. Schoonman,
Eds., Nanostructured Materials: Selected Synthesis Methods, Properties, and Appli-
cations (Electronic Materials: Science & Technology) Kluwer Academic Publishers,
2002.
32. P. Yang, Y. Tu, R. Fan, International Journal of Nanoscience, 1(1) 1, 2002.
33. a) Y. Yu, R. Fan, P. Yang Nano Lett., 2, 83, 2002; b) Y.C. Choi, W.S. Kim, Y.S. Park,
S.M. Lee, D.J. Bae, Y.H. Lee, G.S. Park, W.B. Choi, N.S. Lee, J.M. Kim Adv. Mater.,
12, 746, 2000; c) X.F. Duan, C.M. Lieber, Adv. Mater.,12, 298, 2000; d) T.J. Trentler,
K.M. Hickman, S.C. Geol, A.M. Viono, P.C. Gibbons, W.E. Buhro, Science, 270,
1791, 1995.
34. a) Y. Wu, P. Yang, Chem. Mater., 12, 605, 2000; b) Y. Wu, B. Messer, P. Yang, Adv.
Mater., 13, 1487, 2001; c) Y. Wu, P. Yang, J. Am. Chem. Soc., 123, 3165, 2001; d) M.
Yazara, M. Koguchi, A. Muto, M. Ozawa, K. Hiruma, Appl. Phys. Lett., 61, 2051, 1992.
35. a) J. Stejny, J. Dlugosz, A. Keller, J. Mater Sci., 14, 1291, 1979; b) K.H. Meyer, H.
Mark, Helv. Chim. Acta, 61, 1932, 1937; c) L. Stryer, Biochemistry, 3rd ed., W.M.
Freeman and company, New York, p. 261, 1998.
36. A.A. Kundryavtsev, The Chemistry and Technology of Selenium and Tellurium,
London, pp. 18, 1974.
37. P. Davidson, J.C. Gabriel, A.M. Levelut, P. Batail, Europhys. Lett., 21, 317, 1993.
38. M. Volmer and I. Estermann, Z. Phys., 7, 13, 1921.
39 a) G.W. Sears, Acta. Metall. 3, 367, 1955; b) G.W. Sears, Acta. Metall., 1, 427, 1953.
40. J. Hu, T.W. Odom and C.M. Lieber, Acc. Chem. Res., 32, 435, 1999.
41. a) P. Yang, H. Yan, S. Mao, R. Russo, J. Johnson, R. Saykally, N. Morris, J. Pham,
R. He, H.J. Choi, Adv. Funct. Mater., 12(5), 2002; b) Y.C. Kong, D.P. Yu, B. Zhang,
W. Fang and S.Q. Feng, Appl. Phys. Lett., 78, 4, 2001.
42. C.H. Liang, G.W. Meng, Y. Lei, F. Phillipp, L.D. Zhang, Adv. Mater., 13, 1330, 2001.
43. a) Z.L. Wang, R.P. Gao, J.L. Gole, J.D. Stout, Adv. Mater., 12, 1938, 2000; b) C.H.
Liang, G.W. Meng, G.Z. Wang, Y.W. Wang, L.D. Zhang, S.Y. Zhang, Appl. Phys. Lett.,
78, 3202, 2001.
44. a) Z.W. Pan, Z.R. Dai, Z.L. Wang, Science, 291, 1947, 2001; b) Z.W. Pan, Z.R. Dai,
Z.L. Wang, Appl. Phys. Lett., 80, 309, 2002; c) Z.R. Dai, Z.W. Pan, Z.L. Wang, Solid
State Commun., 118, 351, 2001.
45. R.-Q. Zhang, Y. Lifshitz, and S.T. Lee, Adv. Mater., 15(78), 635, 2003.
46. L.J. Lauhan, M.S. Gudiksen, D. Wang, C.M. Leiber, Nature, 420, 57, 2002.
47. Z.L. Wang, Z.R. Dai, Z.G. Bai, R.P. Gao, J.L. Gole, Appl. Phys. Lett., 77, 3349, 2000.
48. Y.Q. Zhu, W.K. Hsu, W.Z. Zhou, M. Terrones, H.W. Kroto, D.R.M. Walton, Chem.
Phys., Lett., 347, 337, 2001.
49. D. Wang, M. Shao, D. Yu, W. Yu, and Y. Qian, J. Crys. Growth, 254, 487, 2003.
50. Z.L. Wang, Z. Pan, Adv. Mater., 14(15), 2002.
51. H. Yun, R. He, J. Pham, P. Yang, Adv. Mater., 15(5), 402, 2003.
52. C. Ma, D. Moore, J. Li Z.L. Wang, Adv. Mater., 15 (3), 228, 2003.
53. J.Y. Lao, J.G. Wen, D.Z. Wang, Z.F. Ren, Intl. J. Nanosci., 2, 149, 2002.
54. M.H. Huang, S. Mao, H. Feick, H. Yan, Y. Wu, H. Kind, E. Weber, R. Russo, P.
Yang, Science, 292, 1897, 2001.
55. a) E.L. Paradis, A.J. Shuskus, Thin Solid Films, 83, 131, 1976; b) F.C.M. Van, D.E.
Pol, Ceram. Bull., 69, 1959, 1990.
56. a) K. Nashimoto, S. Nakamura, H. Moriyama, Jpn. J. Appl. Phys., 43, 5091, 1995;
b) T. Nagata, T. Shimura, A. Asida, N. Fujimura, T. Ito, J. Crys. Growth, 237, 533, 2002.
57. a) A.J. Freeman, K.R. Poeppelmeier, T.O. Mason, R.P.H. Chang, T.J. Marks, MRS
Bull., 2000, 25, 45, 2000; b) D.S. Ginley, C. Bright, MRS Bull., 25, 15, 2000.
58. a) N. Chubachi, Proc. IEEE, 64, 772, 1976; b) M. Kadota, Jpn. J. Appl. Phys., 36,
3076, 1997.
59. a) G. Heiland, Sensors and Actuators, 2, 343, 1982; b) G. Sberveglieri, S. Groppelli,
P. Nelli, A. Tintinelli, G. Giunta, Sensors and Actuators B, 25, 588, 1995.
60. K. Hara et al., Sol. Energy. Mater. Sol. Cells., 64, 115, 2000.
61. H. Yumoto, T. Inoue, S.J. Lee, T. Sako, K. Nishiyama, Thin Solid Films, 345, 38, 1999.
62. a) C.G. Granqvist, Appl. Phys. A, 57, 19, 1993; b) T. Takada, K. Suzukik, M. Nakane,
Sensor Actuat. B, 13, 404, 1993; c) C. Li, D.H. Zhang, X.L. Liu, S. Han, T. Tang,
J. Han, C.W. Zhou, Appl. Phys. Lett., 82, 1613, 2003; d) D.H. Zhang, C. Li, S. Han,
X.L. Liu, T. Tang, W. Jin, C.W. Zhou, Appl. Phys. Lett., 82, 112, 2003.
63. a) M.H. Huang, Y. Wu, H. Feick, N. Tran, E. Weber, P. Yang, Adv. Mater., 13, 113,
2001; b) Y.W. Wang, L.D. Zhang, G.Z. Wang, X.S. Peng, Z.Q. Chu, C.H. Liang,
J. Crys. Growth, 234, 171, 2003.
64. a) J.J. Wu, S.C. Liu, Adv. Mater., 14, 215, 2002; b) S.C. Lyu, Y. Zhang, H. Ruh, H.J.
Lee, H.W. Shim, E.K. Suh, C.J. Lee, Chem. Phys. Lett., 363, 134, 2002.
65. Y.C. Kong, D.P. Yu, B. Zhang, W. Fang, S.Q. Feng, Appl. Phys. Lett., 78, 4, 2001.
66. B.D. Yao, Y.F. Chan, and N. Wang, Appl. Phys. Lett., 81, 4, 2002.
67. H.-J. Egelhaaf and D. Oelkrug, J. Cryst. Growth, 161, 190, 1996.
68. K. Vanheusden, C.H. Seager, W.L. Warren, D.R. Tallant, and J.A. Voigt, Appl. Phys.
Lett., 68, 403, 1996.
69. a) L. Nilsson, O. Groening, C. Emmenegger, O. Kuettel, E. Schaller, L. Schlapbach,
H. Kind, J.-M. Bonard, and K. Kern, Appl. Phys. Lett., 76, 2071, 2000; b) S.H. Jo,
Y. Tu, Z.P. Huang, D.L. Carnahan, D.Z. Wang and Z.F. Ren, Appl. Phys. Lett., 82,
3520, 2003.
70. D.V. Leff, L. Brandt, and J.R. Heath, Langmuir, 12, 4723, 1996.
71. S.H. Jo, Y. Tu, Z.P. Huang, D.L. Carnahan, D.Z. Wang, Z.F. Ren, Appl. Phys. Lett.,
82, 3520, 2003.
72. C.J. Lee, T.J. Lee, S.C. Lyu, Y. Zhang, H. Ruh, and H.J. Lee, Appl. Phys. Lett., 81,
3648, 2002.
73. R.S. Wagner, W.C. Ellis, Appl. Phys. Lett., 4, 89, 1964.
74. H. Kohno, S. Takeda, Appl. Phys. Lett., 73, 3144, 1998.
75. Definition of the symmetry symbols: for example, 4S
*1
-, the first letter such as 6 or 4
or 2 means the basic 6-, 4-, and 2-fold symmetry of the major core nanowires, respec-
tively. The second letter S or M indicates the single or multiple rows of the secondary
ZnO nanorods. If there is nothing after S or M, all the ZnO nanorods are perpendicular
to the major In
2
O
3
core nanowire. The third symbol
*
means that the secondary nanorods
have an angle with the major core nanowire. The number 1 or 2 after the symbol
*
means that not all the secondary ZnO nanorods, branches have an angle with the core
nanowire.
76. Powder Diffraction File Release 2000, PDF Maintenance 6.0 (International Center
for Diffraction Data, Philadelphia).
77. A.P. Levitt, Ed., Whisker Technology, Wiley-Interscience New York, 1970.
3
Carbon Nanotubes
and Bismuth Nanowires
Mildred S. Dresselhaus, Ado Jorio, and Oded Rabin
CONTENTS
3.1 Introduction
3.2 Carbon Nanotubes
3.2.1 Overview
3.2.2 Nanotube Synthesis
3.2.3 Nanotube Characterization
3.2.4 Raman Spectroscopy
3.2.5 Other Photophysical Techniques
3.2.6 Future Directions for Nanotube Development
3.3 Bismuth Nanowires
3.3.1 Bismuth Nanowire Synthesis
3.3.2 Structural Characterization
3.3.3 Electronic Characterization
3.3.4 Optical Characterization
3.3.5 Future Directions for Bismuth Nanowires
3.3.6 Concluding Remarks
Acknowledgments
Problems
References
3.1 INTRODUCTION
Carbon nanotubes and bismuth nanowires are both model systems for describing the
one-dimensional properties of nanotubes and nanowires and will be discussed in this
context in this chapter. For the case of carbon nanotubes, their one-dimensional aspects
are best seen when they are prepared as single-wall carbon nanotubes (cylinders with
walls one carbon atom in thickness) with approximately 1-nm diameters and lengths
typically of about 10 Pm. Their one-dimensional (1-D) properties are best exemplified
in their 1-D electronic density of states (DOS), because their special physical properties
are closely connected with their unique 1-D electronic DOS. The special properties
associated with bismuth nanowires are closely connected to their unique semimetal-
to-semiconducting transition, which occurs at approximately 50-nm wire diameter,
making available the special properties of bismuth in the form of a semiconductor
with highly anisotropic constant energy surfaces, effective mass components, and
physical properties. This chapter focuses on the preparation and characterization of
these materials.
For both of these systems, simple model calculations can be carried out to explain
most of the unusual properties that are observed in their nanostructured form. These
model calculations can be used to modify the conditions of their physical preparation
to produce desired properties. Characterization of these materials is important to
establish what their intrinsic properties are and how these basic properties can be
used, controlled, or modified for both scientific studies and practical applications.
3.2 CARBON NANOTUBES
3.2.1 OVERVIEW
The special interest in carbon nanotubes stems from their unique structures and
properties, their very small size (as small as ~0.42 nm in diameter), the possibility
for carbon nanotubes to be metallic or semiconducting depending on their geometric
structure, their exceptional properties of ballistic transport, extremely high thermal
conductivity, extremely high optical polarizability, and possibilities of high structural
perfection. Single-wall carbon nanotubes (SWNTs) have only one atomic species
(carbon) and a relatively simple structure (a sheet of regular hexagons of carbon
atoms rolled in a seamless way into a cylinder one atom thick). Single-wall nanotubes
thus provide a benchmark for studying and characterizing new 1-D phenomena
originating from their unique molecular density of states, whereby every structurally
distinct (n,m) nanotube has its own unique and characteristic 1-D density of states,
as discussed below. Many unexpected phenomena that do not occur in the parent
graphite material have been discovered in carbon nanotubes and these discoveries
have energized not only nanotube research, but also carbon and nanoscience research.
Major gaps in basic knowledge remain, with the major obstacle confronting the
carbon nanotube field being the lack of a detailed understanding of the nanotube
growth mechanism. Such understanding is needed so that nanotubes of desired
diameter and chirality can be grown in a monodispersed and controlled way because
of the unusually close connection between nanotube properties (such as their metal-
licity, i.e., whether they are metallic or semiconducting) and their geometric struc-
ture. For this reason, much current research effort is directed toward the synthesis
and characterization of SWNTs.
There are basically three different methods used to grow single-wall carbon
nanotubes (SWNTs): the arc discharge, laser vaporization, and chemical vapor
deposition (CVD) methods.
13
The synthesis methods will be briefly described
subsequently, after the uniqueness of each (n,m) nanotube is further clarified; (n,m)
denotes the structural indices used to characterize SWNTs.
Double-wall carbon nanotubes (DWNTs), consisting of two coaxial cylindrical
SWNTs, one inside the other as in a Russian doll arrangement, provide a prototype
for studying the structures and properties of multiwall carbon nanotubes (MWNTs)
in a quantitative way. Because of the increased stability and durability of DWNTs
and MWNTs relative to SWNTs, the more robust DWNTs and MWNTs are expected
to have more potential for applications, exploiting the exceptional mechanical strength
and stiffness and the very high thermal conductivity of carbon nanotubes.
4
Before
discussing the synthesis and characterization of SWNTs and DWNTs, the unique
aspect of (n,m) nanotubes is presented so that the synthesis and characterization can
be discussed in a proper framework.
Since every structurally distinct (n,m) nanotube can be considered as a distinct
molecule, it is important to provide the framework used to distinguish one nanotube
from another. As illustrated in Fig. 3.1, a single-wall carbon nanotube is geometrically
only a rolled-up graphene sheet. Its structure can be specified or indexed by its
circumferential periodicity ( ), as described using the chiral vector (AAc in Fig. 3.1),
which connects two crystallographically equivalent sites (A and Ac) on a graphene
sheet. In this way, the geometry of each SWNT is completely specified by a pair of
integers (n,m) denoting the relative position of the pair of atoms on a
graphene strip that, when rolled onto each other, form a tube. Here and are basis
vectors of the hexagonal honeycomb lattice, as shown in Fig. 3.1. This chiral vecto-
r also defines a chiral angle T, which is the angle between and the direction
of the graphene sheet. For the tubes with (n,0), called zigzag tubes, and the tubes with
(n,n), called armchair tubes,
5
the translation and rotation symmetry operations are
independent (and these tubes form symmorphic space groups), while for all other (n,m)
tubes, the rotations and translations are coupled (nonsymmorphic space groups).
5
The
assembly of MWNTs follows a Russian doll arrangement with the stacking of tubes
coaxially at an interlayer distance greater than about 0.34 nm, which is the interlayer
separation of turbostratic graphene layers that lack the ABAB . . . interlayer stacking
order found in 3-D graphite. There does not seem to be a correlation between the (n,m)
on one layer and the (nc,mc) of the adjacent layers beyond a preference for minimizing
FIGURE 3.1 Schematic diagram showing a possible wrapping of a two-dimensional
graphene sheet into a tubular form. In this example, a (5,3) nanotube is under construction
and the resulting tube is illustrated on the right.
6
&
C
h
&
& &
C na ma
h
= +
1 2
&
a
1
&
a
2
&
C
h
&
C
h
&
a
1
C
h
a
1
a
2
A
A
1
2
3
1
2
3
4
5
(5,3)
the interlayer separation to be as close as possible to the turbostratic interplanar distance
to maximize the coupling between adjacent SWNTs.
Before the first SWNT was ever synthesized in the laboratory, theoretical
calculations
79
showed that the electronic properties of SWNTs are very sensitive to
their geometric structures. Although graphene (2-D graphite) is a zero-gap semicon-
ductor,
5
theory predicted that carbon nanotubes can serve as metals or semiconduc-
tors with different-size energy gaps, depending very sensitively on the indices (n,m)
or equivalently on the diameter and helicity of the tubes. In addition, each chiral
(n,m) nanotube (0 T 30q) also has a distinct handedness that further defines its
uniqueness.
10
The physics behind this sensitivity of the electronic properties of carbon nanotubes
to their structures can be understood by relating the nanotube electronic structure to
that of the graphene sheet that is rolled to produce the SWNT. The unique band
structure of a graphene sheet shown in Fig. 3.2 has states crossing the Fermi level at
only one point in k-space, denoted by the point K in Fig. 3.2. The quantization of the
electron wave vector along the circumferential direction resulting from the periodic
boundary conditions leads to the formation of energy subbands associated with cutting
lines k
A
PK
1
separated from one another by a distance K
1
2/d
t
. The d
t
is the nanotube
diameter, P is an integer denoting each quantum state in the circumferential direction
for each cutting line, and a quasicontinuum of states k
11
, described by a reciprocal
lattice vector K
2
occurs along the length of such cutting lines.
11
The K
1
and K
2
reciprocal
lattice vectors form the basis vectors of the nanotube Brillouin zone.
An isolated sheet of graphite is a zero-gap semiconductor whose electronic struc-
ture near the Fermi energy is given by an occupied S band and an empty S
*
band.
FIGURE 3.2 (Top) Tight-binding band structure of graphene (a single basal plane of graph-
ite), showing the S and V bands in the main high symmetry directions. (Bottom) Allowed -
vectors of the (5,5), (7,1), and (8,0) tubes (solid lines) mapped onto the graphite Brillouin
zone.
6
M K
20
10
0
10
E
n
e
r
g
y

(
e
V
)

M
M M
K
K K
(5,5) (7,1) (8,0)
&
k
These two bands have a linear dispersion E(k) around the K point in the Brillouin
zone and, as shown in Fig. 3.2, they meet at the Fermi level at the K point. The Fermi
surface of an ideal graphite sheet consists of the six corner K points. The allowed set
of ks in SWNTs, indicated by the lines in Fig. 3.2, depends on the diameter and
helicity of the tube. Whenever the allowed ks include the point K, the system is a
1-D metal with a nonzero density of states at the Fermi level, resulting in a one-
dimensional metal with two linear dispersing bands. When the point K is not included
in the set of allowed states, the system is a semiconductor with different-size energy
gaps depending on both d
t
and T. The d
t
dependence comes from the K
1
2/d
t
relation, and the T dependence comes from the fact that the equi-energies around
the K point are not circles and exhibit trigonally warped shapes
5
. It is important to
note that the states near the Fermi energy for both metallic and semiconducting tubes
are all from states near the K point, and hence their transport and other electronic
properties are related to the properties of the states on the cutting lines, with the
conduction band and valence band states of a semiconducting tube coming from
states along the cutting line closest to the K point.
The general rules arising from symmetry considerations tell us that SWNTs
come in three varieties: armchair (n,n) tubes, which are always metals; (n,m) tubes
with n m 3j, where j is a nonzero integer very tiny-gap semiconductors, that
behave like metals at room temperature; and large-gap semiconductors for which
n m 3j r (Fig. 3.3). As the tube diameter d
t
increases, the band gaps of the
large-gap and tiny-gap varieties decrease with a 1/d
t
and dependence, respec-
tively. The dependence for the tiny gap is due to a curvature effect and depends
on chiral angle, the tiny gap being a maximum for zigzag and zero for armchair
SWNTs. Thus, in Fig. 3.2, a (7,1) tube would be metallic at 300 K, whereas an
(8,0) tube would be semiconducting, while the (5,5) armchair tube would always be
metallic, consistent with whether a cutting line goes through the K point of the 2-
D Brillouin zone. The corresponding electronic density of states for these three
SWNTs are shown in Fig. 3.3. The main point to be emphasized is that the energy
eigenvalues for every (n,m) SWNT are unique, and therefore each (n,m) nanotube
is considered a distinct molecule with potentially different properties. Van Hove
singularities (vHSs), characteristic of the density of states for 1-D systems, are found
for both semiconducting and metallic SWNTs. To the extent that a specific appli-
cation, such as for an electronic device, depends on the distinction between different
(n,m) tubes, the synthesis of SWNTs depends on a high degree of control of the
synthesis process discussed in the next section.
3.2.2 NANOTUBE SYNTHESIS
All three synthesis methods [arc discharge, laser vaporization, and chemical vapor
deposition (CVD)] produce carbon atoms in a hot gaseous form by evaporation
from a condensed phase and these highly energetic carbon atoms then reassemble
themselves to form carbon nanotubes. All methods use nanosize catalytic particles
to induce the synthesis process and to control the diameter of the nanotubes that
are synthesized. For the arc and laser methods, carrier gases are used for transporting
nanotubes from where they are formed in the reactor to where they are collected,
1
2
/d
t
1
2
/d
t
clearing the way for the subsequent growth of SWNTs in the more active reaction
region of the reactor.
13
The arc discharge
14,15
and laser ablation
16
methods were the
first to allow gram quantities of SWNTs to be synthesized, and these methods are
still used today as sources of high-quality samples for specific fundamental studies.
However, the chemical vapor deposition (CVD) method is the method most amenable
for large-scale production of SWNTs. The HiPco (high pressure catalytic decom-
position of carbon monoxide) adaptation of the CVD approach uses floating catalytic
particles and is a common commercial source of SWNTs for both research and
applications use.
17
However, the adaptation of the CVD process based on alcohol
18
produces relatively clean SWNT bundles with fewer residual catalytic particles, and
when operated at relatively low temperatures (e.g., ~650qC), smaller diameter tubes
are favored.
19
Common catalytic particles are Mo, Co, and Fe, but alloys of Co
Mo and FeCo have been found to produce high SWNT yields. Use of CoMo
bimetallic catalysts and a fluidized bed CVD reactor in a CoMo CAT process
20,21
has been especially successful in the efficient production of small-diameter SWNTs.
FIGURE 3.3 Electronic densities of states for the (5,5), (7,1), and (8,0) SWNTs showing
singularities characteristic of 1-D systems. The (5,5) armchair nanotube is metallic for sym-
metry reasons, with a constant energy-independent density of states between the first singu-
larities in the valence and conduction bands. The (7,1) chiral tube displays a tiny band gap
due to curvature effects, but will display metallic behavior at room temperature. The (8,0)
zigzag tube is a large-gap semiconductor.
12
10.0 5.0 0.0 5.0 10.0
(8,0)
(7,1)
(5,5)
Energy (eV)
Rapid progress is being made on nanotube synthesis to increase the control of
the process, steadily narrowing the diameter and chirality ranges of the nanotubes
that are produced, decreasing their defect and impurity content, increasing their
production efficiency and yield, while expanding their functionality. The main direc-
tions in the pursuit of controlled nanotube synthesis include the synthesis of molec-
ular catalytic clusters with atomically well defined sizes and shapes, the development
of mild catalytic synthesis conditions at reduced temperatures, the development of
patterned growth with a high degree of control in nanotube location and orientation,
and the synthesis of complex and organized networks or arrays of nanotubes on
substrates.
13
Much attention is also being given to the synthesis of multiwall carbon nanotubes
(MWNTs) and double-wall carbon nanotubes (DWNTs) because of their attractive-
ness for specific applications where mechanically robust nanostructures are needed.
DWNTs can be considered model systems because of their similarity in physical
properties to MWNTs and their relative simplicity. The synthesis of double-wall
nanotubes can be accomplished either in a direct synthesis process or from a SWNT
whose core has been previously filled with C
60
molecules by a gas phase reactor at
~400qC on an acid-purified open-ended SWNT bundle. The transformation of the
C
60
-filled SWNTs (called peapods) to DWNTs is accomplished by heat treatment
in the temperature range of 800 to 1200qC.
4
Typically the growth of single-wall carbon nanotubes usually involves a catalytic
process, with nanotubes grown in the presence of catalytic particles, but these growth
conditions typically produce a variety of other carbon species, and many other
unwanted constituents. Therefore, much attention has also been given to nanotube
purification from other species, techniques for the characterization of nanotube
purity, and for the preparation and sorting of nanotubes by length, diameter, chirality,
and metallic or semiconducting properties and perhaps other attributes.
22
Chemical
methods are generally used for the removal of metal particles and other unwanted
impurities
2326
and high-temperature heat treatment in vacuum can be used to remove
structural defects such as heptagonpentagon pairs.
26,27
Nanotube synthesis and purification are closely connected from two points of
view. First an improvement in the synthesis process regarding purity and monodis-
persity in tube diameter and chirality would reduce the need for extensive post-
synthesis purification and separation. Second, if efficient and precise purification
and separation techniques can be developed, the stringent limit that the growth
process should produce a single (n,m) species could perhaps be relaxed. The
convergence or combination of the two approaches, that is, the controlled or prefe-
rential growth of a particular (n,m) nanotube, and an effective and easy purification
and separation process
4
could prove powerful in producing nanotubes with well
defined diameters and chiralities.
It is well known that SWNTs can be either semiconducting or metallic, depend-
ing on their geometry (see Section 3.2.1). At present, all known synthesis processes
yield a mixture of semiconducting and metallic SWNTs within a single SWNT
bundle. For DWNTs or MWNTs, the various shells will contain random collections
of semiconducting (S) and metallic (M) constituents with an approximate 2: 1 ratio
expected for S: M tubes, assuming equal a priori probability for occurrence, which
is realized when many different (n,m) constituents are present in the sample. Nan-
otube applications in some cases, such as electronics, require control of the synthesis
process to discriminate between S and M SWNTs, and for circuit applications,
control of diameter and chirality could also be required. Such control of the synthesis
process will be a research challenge of the future.
For the present, effort is also being given to the separation of SWNTs into M-
enriched and S-enriched samples by a variety of processes that preferentially distin-
guish S tubes from M tubes. One of these separation processes is attributed to the
enhanced chemical affinity of the octadecylamine (ODA) surfactant for S SWNTs
dispersed in tetrahydrofuran (THF), rendering MSWNTs more prone to precipitation
when the THF is partially evaporated, leaving behind a supernatant enriched in S
SWNTs.
28
Another separation procedure for enriching M or S SWNTs is based on
ion exchange liquid chromatography of DNA-assisted SWNT solubilization, relying
on the lower negative charge density of DNA-M SWNT hybrids. In this process, M
SWNT-enriched fractions elute before S SWNT-enriched fractions.
29,30
More recently
the separation of M from S SWNTs using alternating current dielectrophoresis in an
aqueous SWNT suspension was achieved, making use of the higher induced dipole
moments of M SWNTs.
31
More recently selective functionalization of M SWNTs
against S SWNTs after addition of 4-chlorobenzenediazonium tetrafluoroborate to
an aqueous suspension of SWNTs was reported, based on the higher activity of
covalent aryl bonds of diazonium with M SWNTs than with S SWNTs.
3234
M/S
separation has also been reported using centrifugation of a surfactant-stabilized
aqueous suspension of SWNTs with the addition of diluted bromine. Here MSWNTs
form more highly structured charge-transfer complexes with bromine than S SWNTs,
thereby yielding an excess of M SWNTs in the supernatant and of S SWNTs in the
sediment.
35
The development of many different separation procedures demands that a simple
and reliable method be developed for the quantitative evaluation of separation effi-
ciency. Several optical techniques widely used in nanotube science can provide such
characterization tools. Optical absorption spectroscopy
36
is one suitable candidate
for the characterization process of the separated samples once the individual peaks
in the absorption spectra are assigned to M and S SWNTs,
37
as discussed in Section
3.2.5. Such measurements are difficult to quantify, because the most intense optical
peak is strongly influenced by exciton effects and because the intensities of all of
the optical absorption peaks are not only influenced by the number density of the
M and S species but by their oscillator strengths for each of the optical transitions.
Resonance Raman spectroscopy (RRS)
38
provides an alternative for the character-
ization of M- and S-enriched aliquots, since M and S SWNTs yield distinct features
in the resonance Raman spectra, as discussed in Section 3.2.4.
39
With the RRS
technique, the intensities of the radial breathing mode peaks are likewise influenced
by the strength of the matrix elements for the Raman intensity. The pertinent matrix
elements for the Raman effect include the electronphoton and electronphonon
matrix elements. In addition, Raman spectra have to be taken at a sufficient number
of laser excitation energies to get a good sampling of the SWNTs present in the
sample. Photoluminescence would be useful in making a determination of how the
diameter distributions change during the separation process, but it is less useful in
evaluating the M and S separation, since M SWNTs show no luminescence and,
furthermore, the presence of M SWNTs in a nanotube sample quenches the lumi-
nescence from S SWNTs by sharply reducing the lifetimes of the excited states of
the S SWNTs through nonradiative decay pathways.
40
3.2.3 NANOTUBE CHARACTERIZATION
There are a number of methods used to characterize the structure and properties of
both individual nanotubes and nanotube bundle samples. In this section, general
characterization approaches are discussed while in Section 3.2.4 the characterization
of SWNTs by Raman spectroscopy is discussed in some detail.
One obvious method to determine whether a SWNT is metallic or semiconduct-
ing would be to measure its electrical conductivity. Such measurements can be done,
but they are not easy to carry out quantitatively because of the small SWNT diameter
and the difficulty in making reproducible electrical contacts. Transport measure-
ments, however, are not useful for specific (n,m) identification.
In practice, electron microscopy has been widely used for structural characteriza-
tion. Scanning electron microscopy (SEM) is routinely used to evaluate the morphology
and the number of remaining catalytic particles and other carbon forms present in an
as-prepared SWNT sample, while an EDAX (energy dispersive x-ray analysis) attach-
ment is typically used for the chemical detection of the noncarbon species present in
the sample and their relative concentrations. Transmission electron microscopy (TEM)
is used to image the lattice planes to evaluate the structural perfection of the sidewalls
of an individual SWNT. Imaging is also used to study the cap structures of SWNTs
and MWNTs, as well as to study the bamboo structures of the inner tubes of MWNTs.
For SWNTs, the TEM is commonly used to obtain electron diffraction patterns, and
from these patterns to determine the (n,m) indices for an individual SWNT.
Scanning probe microscopy (SPM) has been used to characterize isolated
SWNTs spread on surfaces. Atomic force microscopy (AFM) has been widely used
to measure the diameters of individual SWNTs, while low temperature, atomic
resolution scanning tunneling microscopy (STM) images can be used to measure
the chiral angles of individual SWNTs. In the scanning tunneling spectroscopy (STS)
mode, a scanning probe instrument can be used to measure (I/ V) (dI/dV) for indi-
vidual SWNTs, thereby yielding scans of the electronic density of states from which
the (n,m) indices for individual SWNTs can be determined.
Optical techniques, including optical absorption, photoluminescence, and Raman
scattering have been widely used to characterize carbon nanotubes by studying both
their electronic and phonon (lattice vibrational) spectra. These methods have the
advantage of not requiring contacts and in being a very weakly perturbing probe. In
Section 3.2.4, Raman spectroscopy, which is the most widely used spectroscopic tool
for nanotube characterization, is discussed in some detail, while in Section 3.2.5 other
photophysical techniques are briefly reviewed.
3.2.4 RAMAN SPECTROSCOPY
The Raman spectra of SWNTs have been particularly valuable for providing detailed
information for characterizing samples,
38
and several of the features in the Raman
spectra have been used for SWNT characterization, as indicated below. In the first
report of Raman spectra from SWNTs,
38
despite the large number of branches in
the SWNT phonon dispersion relations (Fig. 3.4(a)), the Raman spectra for a SWNT
bundle (Fig. 3.5) exhibited only two dominant features, namely the radial breathing
mode (RBM), which in this trace appears at 186 cm
1
for a laser excitation energy
E
laser
2.41 eV, and the tangential band for vibrations along the surface of the SWNT
appears in the range of 1520 to 1620 cm
1
. Because of the strong connection of this
tangential band to the corresponding mode in 2-D graphite, this higher frequency
band for SWNTs is commonly called the G-band. Other lower intensity features,
seen in Figs. 3.5 and 3.6 and discussed below, also provide important and unique
information about SWNTs.
In this first paper on the Raman effect in SWNTs,
38
the strong and nonmonotonic
dependence of the SWNT Raman spectra on the laser excitation energy E
laser
estab-
lished the Raman scattering to be a resonance process occurring when E
laser
matches
the optical transition energy E
ii
between van Hove singularities in the valence band
and conduction bands (Fig. 3.3). Raman scattering in SWNTs is very important for
the characterization of SWNTs because it relates the photon excitation directly to both
phonon and electron processes. Since the E
ii
value for a particular SWNT is dependent
on its diameter d
t
, so are the Raman spectra that are observed. Because of the very
small diameters of SWNTs (~1 nm), the joint density of states exhibits very large
singularities, with associated large enhancements in Raman intensity, allowing the
FIGURE 3.4 (a) The calculated phonon dispersion relations of an armchair carbon nanotube
with (n,m) (10,10), for which there are 120 degrees of freedom and 66 distinct phonon
branches.
5
(b) The corresponding phonon density of states for a (10,10) nanotube. (c) The
corresponding phonon density of states for a 2-D graphene sheet, for purposes of comparison.
5
1600
1200
800
400
0

(
c
m
1
)
0.02 0.00 0.02 0.00
0.0 0.5 1.0
States/C-atom/cm
1
States/C-atom/cm
1
kT/
(b) (c) (a)
observation of spectra from an individual SWNT that is in strong resonance with
E
laser
.
41
From the point of view of nanotube characterization, the resonance Raman
effect allows the selection for detailed examination of only the few SWNTs in a
sample that are in resonance with a given E
laser
. For example, the RBM for which all
carbon atoms in a given nanotube are vibrating in phase in the radial direction has
a frequency Z
RBM
that is highly sensitive to the nanotube diameter, as discussed
below, thereby providing information on both the mean diameter and the diameter
distribution of a particular SWNT sample.
42
Figure 3.5 indicates that the G-band feature consists of a superposition of two
dominant components, shown at 1593 cm
1
(G
) and at 1567 cm
1
(G
) (see the inset

to Fig. 3.5), although the G
component exhibits a weak diameter dependence. The

G
feature is associated with carbon atom vibrations along the nanotube axis and
its frequency W
G
+ is sensitive to charge transfer from dopant additions to SWNTs
(upshifts in W
G
+ for acceptors, and downshifts for donors). The G
feature, in contrast,
is associated with vibrations of carbon atoms along the circumferential direction of
the nanotube, and its line shape is highly sensitive to whether the SWNT is metallic
FIGURE 3.5 Experimental Raman spectrum taken with 514.5 nm (2.41 eV) laser excitation
from a SWNT bundle sample with a diameter distribution d
t
1.36r20 nm. The inset shows
an expanded version of the spectra in the 1450 to 1700 cm
1
range, fit by a sum of Lorentzians.
38
116
186
376
754 857
1347
1736
1567
1593
1450 1500 1550
1550
1567
1600
1606
1650
1526
1593
Purified SWNT
500 1000 1500 2000
Wavenumbers
Wavenumbers
Purified SWNT
R
a
m
a
n

i
n
t
e
n
s
i
t
y
R
a
m
a
n

i
n
t
e
n
s
i
t
y
(BreitWignerFano line shape) or semiconducting (Lorentzian line shape).
43,44
All
in all, there are six modes contributing to the G-band, with two each having A, E
1
,
and E
2
symmetries, and each symmetry mode can be distinguished from the others
by its behavior in polarization-sensitive Raman experiments.
45
The features of
highest intensity are the G
and G
features of A symmetry, and these components

are normally used to characterize carbon nanotubes with regard to their metallicity
and involvement in charge transfer. Polarization studies on SWNTs are best done
at the single-nanotube level.
46
Also commonly found in the Raman spectra in SWNT bundles are the D-band
(with Z
D
at 1347 cm
1
in Fig. 3.5), stemming from the disorder-induced mode in
graphite, and its second harmonic, the band (Fig. 3.6) occurring at ~2Z
D
. Both
the D-band and the -band are associated with a double resonance process,
47
and
are sensitive to the nanotube diameter and chirality. The large dispersions of the D-
band and -band features have been very important in revealing detailed informa-
tion about the electronic and phonon properties of SWNTs as well as of the parent
material graphite.
4850
Because of the sharp van Hove singularities occurring in carbon nanotubes with
diameters less than 2 nm, the Raman intensities for the resonance process can be
so large that it is possible to observe the Raman spectra from one individual SWNT,
41
as shown in Fig. 3.6, where the differences in the G-band spectra between semi-
conducting and metallic SWNTs can be seen at the single-nanotube level. Because
of the trigonal warping effect, whereby the constant energy contours in reciprocal
k space for 2-D graphite are not circles but show three-fold distortion, every (n,m)
carbon nanotube has a different electronic structure and a unique density of states
(Fig. 3.3). Therefore the energies E
ii
of the van Hove singularities in the joint density
FIGURE 3.6 Raman spectra from metallic (top) and semiconducting (bottom) SWNT at the
single nanotube level using 785-nm (1.58-eV) laser excitation, showing the radial breathing
mode (RBM), D-band, G-band, and Gc band features, in addition to weak double resonance
features associated with the M-band and the iTOLA second-order modes. Insets on the left
and the right show, respectively, atomic displacements associated with the RBM and G-band
normal mode vibrations. The isolated carbon nanotubes are sitting on an oxidized silicon
substrate that provides contributions to the Raman spectra denoted by asterisks and are used
for calibration purposes.
RBM
Metallic
RBM
G-band
0 500 1000 1500 2000 2500 3000
Semiconducting
iTOLA
785 nm
D
D
G
G
G'
G'
M
Si/SiO
2
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
Frequency (cm
1
)
G
G
G
of states for each SWNT (as shown in Fig. 3.4(b) and calculated, for example, by
the tight binding approximation) are different, as shown in Fig. 3.7, where the E
ii
(i 1, 2, 3, ...) values for all (n,m) nanotubes in the diameter range up to 3.0 nm
are shown for E
ii
up to 3.0 eV.
51
In this so-called Kataura plot (Fig. 3.7), we see
that for small-diameter SWNTs (d
t
1.7 nm) and for the first few electronic
transitions for semiconducting and metallic SWNTs, the E
ii
values are arranged in
bands, whose widths are determined by the trigonal warping effect
49
so that if there
were no trigonal warping, the spread in energy at constant diameter would go to
zero. A general chiral semiconducting SWNT will have van Hove singularities in
the bands, while a chiral metallic SWNT will have two vHSs in
each of the bands. The tight binding approximation provides predicted
values for the E
ii
energies for the vHSs for SWNTs with diameters in the 1.1-2.0-
nm range (to an accuracy of better than 20 meV).
52
This Kataura plot can be used
to estimate the appropriate laser energy that could be used to resonate with a
particular (n,m) SWNT.
The Raman effect furthermore provides a determination of E
ii
values, either
by measurement of the relative intensities of the radial breathing mode for the
Stokes (phonon emission) and anti-Stokes (phonon absorption) processes or by
FIGURE 3.7 Calculated
51
energy separations E
ii
between van Hove singularities i in the 1-D
electronic density of states of the conduction and valence bands for all possible (n,m) values
versus nanotube diameter in the range 0.4 d
t
3.0 nm, using a value for the carboncarbon
energy overlap integral of J
0
2.9 eV and a nearest neighbor carboncarbon distance a
CC
0.142 nm for making this plot.

49,58
Semiconducting (S) and metallic (M) nanotubes are
indicated by crosses and open circles, respectively. The subscript i 1 denotes the index of
the lowest energy of a singularity in the joint density of states.
0.4 0.9 1.4 1.9 2.4 2.9
0.00
1.00
2.00
3.00
E
i
i
(
d
t
)

(
e
V
)
d
t
(nm)
0
= 2.90 eV E
S
11
E
S
22
E
M
11
E
S
33
E E E
S S S
11 22 33
, , , . . .
E E
M M
11 22
, , . . .
measurements of the RBM Raman intensity for the Stokes process for a particular
(n, m) SWNT relative to the intensity for many SWNTs measured under similar
conditions.
39
The frequency of the RBM has been used to determine the diameter
of an isolated SWNT sitting on an oxidized Si surface, using the relation Z
RBM
(cm
1
) 248/d
t
(nm), which is established by measurements on many SWNTs.
Moreover, from a knowledge of the (E
ii
, d
t
) values for an individual SWNT (Fig. 3.7),
the (n, m) indices for that SWNT can be determined from the Kataura plot. Reso-
nance Raman spectroscopy at the single-nanotube level can not only determine the
(n, m) indices from study of the RBM frequency and intensity, but can also deter-
mine E
ii
values for individual SWNTs to about 10-meV accuracy, using a single
laser line or to about 5-meV accuracy using the Stokesanti-Stokes intensity
analysis method.
39
Resonance Raman measurements with a tunable laser system
can give E
ii
values with an accuracy of r3 meV.
53,54
The precision for this assign-
ment using Stokes versus anti-Stokes Raman measurements depends on the deter-
mination of the shape of the resonance window (Raman intensity as a function of
excitation laser energy E
laser
). The resonance window could change for samples
with different environments and with different sample preparation methods, and
systematic work is still needed to increase the precision in determining E
ii
. The
ability to use a gate,
55
an externally applied potential,
56
or a tunable laser
53,54,57
to move
the van Hove singularity for an individual SWNT into and out of resonance with
the laser offers great promise for future detailed studies of the 1-D physics of single-
wall carbon nanotubes using resonance Raman spectroscopy. The ability to char-
acterize individual SWNTs for their (n, m) identification allows the determination
of the dependence on nanotube diameter and chirality of many physical phenomena
in SWNTs that can be measured at the single-nanotube level.
Near-field microscopy has also proven to provide a powerful tool for examining
the Raman spectra on a spatially resolved basis (to 20-nm resolution), thus providing
detailed information about the effects of specific defects on the vibrational spectra.
The disorder-induced D-band provides a sensitive probe of symmetry-breaking
agents such as defects in carbon nanotubes.
59,60
Raman spectra of carbon nanotubes, particularly at the single-nanotube level,
have been especially rich. Because of the simplicity of the geometrical structure
of nanotubes, detailed analysis of the Raman spectra have yielded much information
about the phonon dispersion relations, such as information about the trigonal
warping effect for phonons. Such information was not yet available for 2-D graphite,
but can now be studied in nanotubes because of their one-dimensionality.
50
Because
of the close coupling between electrons and phonons under resonance conditions,
Raman spectra have also provided valuable and detailed information about the
electronic structure (such as an evaluation of the magnitude of the trigonal warping
effect for electrons
48
), thus yielding valuable information about the electronic
structure that can be used in the characterization of nanotube samples. At present,
the use of many laser lines (over 50) provides the capability of carrying out Raman
spectroscopy measurements with a quasicontinuous excitation energy source.
54,57
This capability promises to provide a means to access essentially all the SWNTs
in a given sample, thereby providing a powerful new characterization technique for
SWNTs.
3.2.5 OTHER PHOTOPHYSICAL TECHNIQUES
Optical absorption measurements on carbon nanotube samples generally show
broad peaks
61
corresponding to and with decreasing intensity as the
energy increases, as expected from the f -sum rule for optical matrix elements.
62
These peaks are characteristic of SWNTs and provide one of the most important
characterization techniques to distinguish nanotubes from other carbonaceous spe-
cies in an unpurified sample containing nanotubes in the presence of other carbon-
aceous species. The characterization measurement is carried out by comparing the
absorption intensity of the band relative to the background absorption in the
same range of the infrared spectra. The peak is selected because it is relatively
strong and is not expected to be strongly perturbed by excitonic effects, as is the
peak. Optical absorption spectra are not only used for the characterization of such
mixed samples but as an aid in the purification of such samples to isolate the
nanotubes from other species.
22,63
Although optical absorption techniques can be used to characterize almost any
type of sample containing carbon nanotubes, optical emission as a characterization
tool requires specially prepared nanotube samples with sufficiently long carrier
lifetimes in the excited state, so that the emission process has sufficient intensity to
be observable,
40
as discussed in the next paragraph.
Time-domain photoemission studies using fast optics provide information about
the lifetime and electronphonon matrix elements for the excited states. Time-
domain studies on SWNT bundles have been performed, giving a measure of the
lifetime of photoexcited charge carriers in SWNT ropes as a function of electron
energy (Fig. 3.8).
64,65
The lifetimes of electrons excited to the S
*
bands were found
FIGURE 3.8 Time dependence of the photoelectron intensity during and after femtosecond
laser generation of hot electrons in SWNT bundles.
65
Electrons monitored by this trace are
30 meV above the Fermi level. The fast and slow components, respectively, correspond to
internal thermalization of the laser-heated electron gas and to its equilibration with the lattice,
i.e., by cooling.
E E
S S
11 22
, , E
M
11
E
S
22
E
S
22
E
S
11
0 2 4 6
1
10
100
Electron gas cooling:
e-ph interactions
Laser-heated electron
thermalization: e-e interactions
(E E
F
) = 30 meV
Pump-probe delay (ps)
C
h
a
n
g
e

o
f

p
h
o
t
o
e
l
e
c
t
r
o
n

i
n
t
e
n
s
i
t
y
to decrease continuously from 130 fs at 0.2 eV down to less than 20 fs at energies
above 1.5 eV with respect to the Fermi level.
64
These short lifetimes lead to a
significant lifetime-induced broadening of the characteristic vHSs in the density of
states, so that the optical effects related to the confinement of electrons into vHSs
are stronger for the lower energy vHSs in the joint density of states. Experimental
results show very similar excited state decay times for graphite and SWNT bundles,
suggesting that electronelectron scattering of photoexcited carriers in SWNT bun-
dles may lead to a rapid charge transfer between different tubes, thus allowing excited
electron hole pairs in semiconducting SWNTs to relax almost as rapidly as those in
metallic tubes. This result probably explains why photoluminescence is not observed
for SWNT bundles.
40
The presence of conduction electrons in metallic tubes provides
an effective nonradiative decay channel for the decay of photoexcited electrons. Thus
to obtain photoluminescence spectra, the SWNTs contained in SWNT bundles need
to be isolated from one another to reduce the nonradiative decay channel, and this
has been done by the use of various species to wrap around individual SWNTs within
a bundle to isolate individual SWNTs from each other. Some wrapping agents that
have been effectively used include sodium dodecyl sulfate (SDS)
40,66
and single-
stranded guaninethyamine selected DNA chains (ss-GTDNA).
29
The observation of photoluminescence (PL) from isolated SWNTs has made
possible the observation of the energy gap for semiconducting SWNTs, and
interesting 3-D plots can be constructed (Fig. 3.9) showing strong intensity peaks
for excitation of the electrons at and emission at , where each peak is related
to one specific (n,m) SWNT. From this intensity plot in Fig. 3.9, a plot of the E
11
and E
22
electronic transition energies for many (n,m) SWNTs can be constructed.
66
By using an empirical expression for E
ii
as a function of nanotube diameter and
chiral angle for isolated SWNTs grown by the HiPco process, wrapped with SDS
surfactant
67
and dispersed in aqueous solution, an assignment of the PL peaks with
specific (n, m) values has been made. This assignment is based on the identification
of families with (2n m) constant, which becomes increasingly evident as the
SWNT diameter decreases and curvature effects become more important. The widths
of the PL peaks are ~25 meV, corresponding to room temperature thermal
energy.
37,40,66,67
The strength of the PL characterization technique is that all the (n,m)
semiconducting tubes in the sample are probed at once so that, in principle, their
relative concentrations could be determined (if the matrix elements and lifetimes
for the (n,m) emissions were known theoretically) from the measured intensities of
each PL peak.
The limitation for the photoluminescence method is related to systems where
nonradiative electron-hole recombination can occur, so that light emission from
metallic SWNTs or SWNT in bundles cannot be observed. For such samples,
resonance Raman experiments at the single-nanotube level could be alternatively
used. Raman measurements on SDS-wrapped SWNTs in solution yield good agree-
ment with values obtained by PL measurements
54
within the accuracy of the PL
measurements, which can only be made to 25 meV because of the broad line widths.
In contrast, by using many laser lines, the accuracy of the values is much greater
(to 5 meV) as determined by resonance Raman spectroscopy, through measurement
of both the Stokes and anti-Stokes RBM features, as discussed in Section 3.2.4.
E
S
11
E
S
22
E
S
11
E
S
22
E
S
22
The discrepancies observed among various types of nanotube samples for the
and transition energies (up to 100 meV) show that the electronic energy levels that
we use to distinguish one type of (n,m) SWNT from another are, in fact, sensitive to
the different wrapping agents used, and to the different environments of the SWNTs
(freely suspended, isolated SWNTs on various oxide substrates, in aqueous or organic
solvents, etc.).
54
Such phenomena are currently under investigation.
3.2.6 FUTURE DIRECTIONS FOR NANOTUBE DEVELOPMENT
A large amount of information at the isolated SWNT level has provided accurate
data concerning the structures and properties of single-wall carbon nanotubes and
contributed in an important way to the development of more complete theoretical
models to describe nanotube physics. Much effort must now focus on using these
photophysics studies to characterize carbon nanotube systems in detail, because for
such a nanometer scale system, the massless photon turns out to be a powerful probe
for nanotube characterization without strongly disturbing this nanoscale system. For
this reason we can expect spectroscopy to continue to serve as an important tool for
characterizing nanotube properties reliably.
Although sample purity can be investigated using the D/G band intensity ratio
in the Raman spectra from SWNTs, the development of systematic work is needed
FIGURE 3.9 Fluorescence intensity versus excitation and emission wavelengths for SWNTs
in an SDS suspension.
67
Since each peak corresponds to the absorption/emission of a single
(n,m) SWNT, the intensity of each peak provides a measure of the relative concentrations of
each (n,m) SWNT in the sample.
300
900
1000
1100
1200
1300
1400
1500
400
500
600
700
800
E
x
c
ita
tio
n
w
a
v
e
le
n
g
th
(n
m
)
E
m
i
s
s
i
o
n

w
a
v
e
l
e
n
g
t
h

(
n
m
)
E
S
22
E
S
11
for quantitative information about sample purity using Raman spectroscopy. A quan-
titative procedure for the evaluation of the carbonaceous purity of bulk quantities of
as-prepared SWNT soot can be obtained by the utilization of solution-phase near-
infrared spectroscopy.
68
However, for defect characterization at the single-nanotube
level, development of the physics related to the defect-induced Raman D-band feature
is needed. Such a development will be important for the characterization of more
advanced nanotube-based devices, where any junction, kink, doping additive, or
wrapping agent will be seen by the Raman effect as a defect, and will affect device
performance differently. The challenge will be to see how each of these different
types of defects can provide a different and distinguishable footprint in Raman
spectroscopy. By using polarized light and exploiting selection rules at the single-
nanotube level and by using near-field Raman spectroscopy, a variety of challenging
and promising experiments can be carried out for the development of the next
important phase of carbon nanotube photophysics characterization.
For applications of SWNTs as electronic devices it is essential to have a source
of inexpensive, pure single-phase material, preferably with the same (n,m) indices.
Although progress has been made in both the synthesis of a more uniform carbon
nanotube product and in the separation of metallic from semiconducting nanotubes,
22
the synthesis technology is not sufficiently developed for large-scale nanotube elec-
tronics applications for commercial products. Development of (n,m) selective syn-
thesis is the most challenging present goal for the development of carbon nanotube
science and applications.
3.3 BISMUTH NANOWIRES
Nanowires are especially attractive for nanoscience studies as well as for nanotech-
nology applications. Nanowires, compared to other low-dimensional systems, have
two quantum confined directions, while still leaving one unconfined direction for
electrical conduction. This allows nanowires to be used in applications where
electrical conduction, rather than tunneling transport, is required. Because of quan-
tum confinement effects, nanowires in the limit of small diameters are expected to
exhibit significantly different optical and electrical properties from their bulk 3-D
crystalline counterparts and this is very much the case for bismuth nanowires. Their
increased surface area, diameter-dependent band overlap and bandgap, and
increased surface scattering for electrons and phonons are only some of the ways
in which nanowires differ from their corresponding bulk materials. However the
diameters of typical bismuth nanowires are large enough (!5 nm) to have crystal
structures closely related to their parent material, thereby allowing theoretical
predictions about their properties to be made on the basis of an extensive literature
relevant to bulk bismuth properties. Not only do bismuth nanowires exhibit many
properties that are similar to, and others that are distinctly different from, those of
their bulk counterparts, but they also have the advantage from an applications
standpoint that some of their materials parameters that are critical for certain
physical properties can be independently controlled in their nanowire form, but
not in bulk bismuth, such as, for example, thermal and electrical conductivity.
Certain properties can also be enhanced nonlinearly in small-diameter nanowires
by exploiting the singular aspects of the 1-D electronic density of states. Bismuth
nanowires provide a promising framework for applying the bottom-up approach
69
for
the design of nanostructures for nanoscience investigations and for potential appli-
cations as, for example, for thermoelectricity. It is the aim of this section to focus
on bismuth nanowire properties that differ from those of bulk bismuth, with an eye
toward possible applications that might emerge from the unique properties of these
nanowires and from future discoveries in this field.
3.3.1 BISMUTH NANOWIRE SYNTHESIS
Bismuth (Bi) nanowires in principle can be prepared by many techniques
70
and it is
expected that the synthesis of Bi nanowires by various novel techniques will become
an even more active current research area than it has been. Most of the presently
available physical property studies of bismuth nanowires have been made on nano-
wires grown in porous anodic alumina templates. A variety of other templates are
available for use in growing bismuth nanowires, such as those prepared from
nanochannel glass, block copolymers, and ion track-etched polymer and mica films.
Using porous anodic alumina templates (Fig. 3.10), highly crystalline bismuth
nanowires with diameters in the 7- to 200-nm range and lengths up to 10 Pm have
been prepared with parallel alignment and a high packing density (~10
11
nanow-
ires/cm
2
), which is favorable for a number of applications.
71
Nanochannel Vycor
glass has been used to prepare Bi nanowires down to 4-nm diameter,
72
and the
nanowires thus prepared have been useful for establishing the value of the wire
diameter below which localization effects become important in electron transport.
However, the random organization of the pores, and consequently of the wires filling
the template pores, would limit the range of applications of such nanowire arrays.
Polymer-based templates might be desirable because of their favorable properties
for certain applications (such as low thermal conductivity of the host material, which
is desirable for thermoelectric applications), but because of their limitations for use
FIGURE 3.10 Porous anodic alumina templates showing (a) a schematic and (b) an SEM
picture of the top surface.
75
The pores of these templates are filled to form nanowires.
Porous
anodic
alumina
Pore channel
Aluminum
(a) (b)
at elevated temperatures, the filling of polymer-based templates would likely have
to be done near room temperature by electrochemical means.
73,74
The filling of the porous anodic alumina templates (Fig. 3.10) to form Bi
nanowire arrays has been accomplished by at least three methods, including pressure
injection, vapor growth, and electrochemical methods. Most of the properties mea-
surements have been made with Bi nanowires prepared by pressure injection and
vapor growth, although the electrochemical method is by far the most versatile and
is the most logical candidate for scale-up for commercial applications.
The pressure injection technique can be employed for fabricating bismuth
nanowires because of the low melting point of this material (271.4qC). In the high-
pressure injection method, the nanowires are formed by pressure injecting the
desired material in liquid form into the evacuated pores of the template. Due to
the heating and the pressurization processes, the templates used for the pressure
injection method must be chemically stable and be able to maintain their structural
integrity at about 50 K above the melting point of bismuth at the injection pressure.
Anodic aluminum oxide films and nanochannel glass are two typical host materials
that have been used as templates in conjunction with the pressure injection filling
technique.
7678
The pressure P required to overcome the surface tension for the liquid material
to fill the pores with a diameter d
W
is determined by the Washburn equation, P
4JcosT/d
W
, where J is the surface tension of the liquid, and T is the contact angle
between the liquid and the template.
79
To reduce the required pressure and to
maximize the filling factor, some surfactants are used to decrease the surface tension
and the contact angle. For example, it is found that the introduction of a small amount
of Cu in the Bi melt can facilitate the filling of the anodic alumina pores with liquid
Bi and increase the number of nanowires that are prepared.
80
A typical pressure used
for the preparation of bismuth nanowires is 0.3 to 1.5 kbar.
80,81
Nanowires produced
by the pressure injection technique usually possess high crystallinity and a preferred
crystal orientation along the wire axis.
The electrochemical deposition technique has attracted increasing attention as
a promising alternative for fabricating nanowires because of its versatility and
ability to produce bismuth nanowires of smaller diameter. Traditionally electro-
chemistry has been used to grow thin films on conducting surfaces. Since electro-
chemical growth is usually controllable in the direction normal to the substrate
surface, this method can be readily extended to fabricate both quantum wire- and
quantum dot-based nanostructures, if the deposition is confined within the pores
of an appropriate template. In the electrochemical methods, a thin conducting metal
film is first applied on one side of the porous membrane to serve as the working
electrode for electroplating. The length of the deposited nanowires can be controlled
by varying the duration of the electroplating process. This method has been used
to synthesize a wide variety of electrically conducting nanowires including bismuth
nanowires.
82
In the electrochemical deposition process, the chosen template must be chemi-
cally stable in the electrolyte during the electroplating process. Cracks and defects
in the templates are detrimental to nanowire growth because electrochemical depo-
sition will then occur primarily in the more accessible cracks, thus leaving most of
the nanopores unlled. Particle track-etched mica lms and polymer membranes are
typical templates used in the simple DC electrolysis preparation method. To use
anodic aluminum oxide lms for DC electrochemical deposition, the insulating
barrier layer that separates the pores from the bottom aluminum substrate has to be
removed, and a metal lm is then evaporated onto the back of the template mem-
brane.
83
The electrochemical lling of anodic alumina templates has been carried
out on both Bi and Bi
2
Te
3
.
8486
It is also possible to employ an AC electrodeposition method in lling anodic
alumina templates without the removal of the barrier layer by utilizing the rectifying
properties of the oxide barrier. In AC electrochemical deposition, although the applied
voltage is sinusoidal and symmetric, the current is greater during the cathodic half-
cycles, making deposition dominant over the etching, which occurs in the subsequent
anodic half-cycles. Since no rectication occurs at defect sites, the deposition and
etching rates are equal, and no material is deposited. Hence the difculties associated
with cracks are avoided. Pulse current electroplating techniques have been used for
the synthesis of Bi nanowires
87,88
and they appear to be advantageous for the growth
of crystalline wires generally because the metal ions in the solution can be regenerated
between the electrical pulses and, therefore, more uniform deposition conditions can
be produced for each deposition pulse.
One advantage of the electrochemical deposition technique is the possibility of
fabricating multilayered structures within nanowires. By varying the cathodic poten-
tials in the electrolyte, which contains two different kinds of ions, layers of different
compositions can be controllably deposited. This electrodeposition method provides
a low-cost approach to preparing multilayered 1-D nanostructures. One disadvantage
of the electrochemical deposition approach for applications is that the Bi nanowires
fabricated by the electrochemical process are often polycrystalline, with no preferred
crystal orientations. Progress has been made in improving the crystallinity of Bi
nanowires prepared by the electrochemical route by careful control of the deposition
parameters.
75,8991
Vapor deposition is a convenient method for lling the template pores with bismuth
and is usually capable of preparing smaller diameter (20 nm) nanowires than pressure
injection methods produce because it does not rely on high pressure to overcome the
surface tension involved in inserting the bismuth into the pores of the template. In the
physical vapor deposition technique, the bismuth is rst heated to produce a vapor,
which is then introduced through the pores of the template, and the template is subse-
quently cooled to solidify the bismuth nanowires. Using a specially designed experi-
mental setup,
92
nearly single-crystal Bi nanowires in porous anodic alumina templates
with pore diameters as small as 7 nm have been synthesized, and these Bi nanowires
were found to possess a preferred crystal growth orientation along the wire axis, similar
to the Bi nanowires prepared by pressure injection.
92
Once the bismuth nanowires are grown in a template, it is possible to etch away
the template by appropriate chemical treatments to gain access to the individual
nanowires. An example of such free-standing bismuth nanowires on a silicon sub-
strate is shown in Fig. 3.11.
Although porous alumina templates provide a convenient method for preparing
arrays of nanowires for study and demonstration purposes, these arrays have some
severe disadvantages for large-scale applications. Porous anodic alumina is a brittle
ceramic film grown on a soft aluminum metal substrate. Great care needs to be
exercised in the preparation of the aluminum substrate and in the manipulation of the
anodic film to produce pure defect-free porous anodic alumina films that are required
in order to achieve uniform filling of the pores with the nanowire material. Between
the porous layer and the aluminum substrate lies a continuous and dense alumina
barrier layer that prevents the direct physical and electrical contact between the pore
(or nanowire) and the substrate. It is therefore of great interest to develop methods
to grow porous anodic alumina films on rigid substrates of technological relevance
and to establish contact of both ends of the pores (or of the nanowires) while the film
is still attached to the substrate. Such technology would allow integration of addres-
sable arrays of parallel nanowires as active components in electronic, thermoelectric,
optical, field emission, and sensing devices.
Techniques have recently been developed to grow arrays of bismuth nanowires
on the surface of a silicon wafer coated with a porous anodic alumina film.
93
A
conducting layer under the porous film serves as the working electrode for the
electrochemical growth of the nanowire array. An example of bismuth nanowires
grown electrochemically on the surface of a silicon wafer is shown in Fig. 3.12.
75,93
Unlike free-standing porous alumina films containing bismuth nanowires, the films
grown on a silicon wafer are robust enough to allow growth on large areas of the
silicon wafer. Furthermore, patterned porous films and patterned electrodes can be
used to control the location and the shape of the nanowire array using conventional
microelectronics fabrication techniques.
3.3.2 STRUCTURAL CHARACTERIZATION
Thus far, most of the emphasis on the structural characterization of bismuth nano-
wires has been in terms of degree of crystallinity and the crystalline orientation
of the nanowire axis. The reason for this focus on crystallinity stems from the
FIGURE 3.11 Bismuth nanowires on a silicon wafer exposed by etching a porous anodic
alumina film.
extraordinarily large anisotropy of the effective mass components and consequently
of the constant electronic energy surfaces of crystalline bismuth.
71
This anisotropy
is responsible for the exceptional electronic properties of bismuth and for the interest
in bismuth nanowires both as model systems for studying 1-D physics phenomena
and for possible future practical applications.
Scanning electron microscopy (SEM) produces images of bismuth nanowires
down to length scales of ~10 nm, and provides valuable information regarding the
structural arrangement, the spatial distribution of the nanowires, the filling factor of
the nanowires in space, the morphological features of individual nanowires, and arrays
of nanowires. Since many of the studies of bismuth nanowires focused on arrays of
bismuth nanowires in templates, SEM has also provided valuable characterization of
the templates used to grow the nanowires to determine the pore diameters and
diameter distribution in the template arrays, the packing fraction, and the filling factor,
as illustrated in Fig. 3.10, in which the top view of a typical empty anodic alumina
template is shown. After the template has been filled with bismuth (by either liquid
phase pressure injection, vapor phase injection, or electrochemical filling), the filling
factor can be determined utilizing the very high contrast of the secondary electron
emission (electron absorption) of the bismuth nanowire as compared to the very low
interaction of the electron probes with the empty pores of the template.
Transmission electron microscopy provides a wide range of characterization
information about bismuth nanowires including their crystal structure, crystal quality,
grain size, and crystalline orientation of the nanowire axis. When operating in the
diffraction mode, selected area electron diffraction (SAED) patterns can be used to
determine the crystal structures of nanowires. The amount of crystallinity of indi-
vidual nanowires is evaluated from lattice fringe images, such as the one shown in
Fig. 3.13(a). Here we can see from the regularity of the lattice fringes that pressure
injection produces bismuth nanowires with a high degree of crystallinity. The crystal-
line quality of bismuth nanowires obtained by vapor phase growth is also very high.
FIGURE 3.12 Cross-section of an array of bismuth nanowires grown on a platinum-coated
silicon wafer, allowing more robust growth of bismuth nanowires in porous anodic alumina
films over large areas on a silicon wafer.
93
To establish the crystalline direction of the nanowire axis, SAED measurements are
made at the individual nanowire level (Fig. 3.13(b)). Such SAED patterns can also
be used to establish the degree of crystallinity along a single nanowire by taking
multiple SAED spectra at different points along the nanowire axis, and SAED
measurement of several nanowires provides information about the distribution of
the crystalline orientations.
X-ray diffraction (XRD) measurements on an array of aligned bismuth nanowires
inside their templates can be used to acquire information about the crystallographic
orientation of the nanowires within a single template, as shown in Fig. 3.14. Note
the high degree of crystallinity with >80% crystalline orientation of the nanowire
along a specic nanowire direction. The large diameter (>95 nm) nanowires show
alignment predominantly along a (101) axis, while the smaller diameter nanowires
(40 nm) show predominant alignment along an (012) axis, suggesting a wire
diameter-dependent crystal growth mechanism. On the other hand, 30-nm Bi nanow-
ires produced using a much higher pressure of >1.5 kbar show a different crystal
orientation of (001) along the wire axis,
81
indicating that the preferred crystal ori-
entation may also depend on the applied pressure, with the most dense packing
direction along the wire axis for the highest applied pressure. The crystallinity and
orientation of vapor-grown bismuth nanowires in the alumina templates are similar
to those of nanowires prepared by pressure injection. Although early work on the
electrochemical preparation of Bi nanowires produced fewer crystalline nanowires
than the vapor-grown or pressure injection methods, subsequent efforts
8991
have pro-
duced aligned and single-crystal nanowires, with crystal orientations along the wire
axes depending on the preparation conditions. X-ray energy dispersion (EDS) anal-
ysis also provides information on the impurity content of the bismuth nanowires.
FIGURE 3.13 (a) HRTEM image of a free-standing individual Bi nanowire with a diameter
of 45 nm. The inset, shown in an enlarged form on the right, highlights the lattice fringes of
the nanowire, indicating high crystallinity. (b) Selected area electron diffraction (SAED)
pattern of the same nanowire in the TEM image.
94
10 nm
(a) (b)
XRD, EDS and TEM characterizations of Bi
1x
Sb
x
alloy nanowires for x 0.15
prepared by the pressure injection method
94
show good compositional homogeneity
along the nanowire axis and good crystallinity of the alloy nanowires, with the same
preferred crystalline alignment of the nanowire axis, as is the case of pure bismuth
nanowires. On the other hand, a gradient in alloy composition along the nanowire
axis was determined by EDS in 200-nm Sb-rich Bi
1x
Sb
x
nanowires electrochemi-
cally deposited in porous alumina templates.
75
It is signicant that the SAED patterns
(such as in Fig. 3.13(b)) noted for individual nanowires by electron diffraction are
in good agreement with the x-ray diffraction patterns taken on large arrays of
nanowires, as in Fig. 3.14.
The high resolution of the TEM also allows for the investigation of the surface
structures of bismuth nanowires that form native surface oxides when exposed to
oxidative conditions (i.e., atmospheric air). Knowledge of the surface composition and
structure is critical for transport applications of bismuth and related nanowires, as, for
example, for thermoelectric cooling or thermal management applications.
71
In an
attempt to control the formation of these oxide surface layers, the surface structures of
bismuth nanowires have been studied by using an in situ environmental TEM chamber
that allowed TEM observations to be made while different gases were introduced or as
the sample was heat treated at various temperatures, as illustrated in Fig. 3.15. This
gure shows the high resolution TEM images of a Bi nanowire with an oxide coating
FIGURE 3.14 XRD patterns of bismuth/anodic alumina nanocomposites with average bis-
muth wire diameters of (a) 40 nm, (b) 52 nm, and (c) 95 nm.
78
The Miller indices corre-
sponding to the lattice planes of bulk Bi are indicated above the individual peaks. The
majority of the Bi nanowires are oriented along the [10 1] and [01 2] directions for d
W
60 nm and d
W
50 nm, respectively.
78
The existence of more than one dominant orientation
in the 52-nm Bi nanowires is attributed to the transitional behavior of intermediate diameter
nanowires as the preferential growth orientation is shifted from [10 1] to [01 2] with
decreasing d
W
.
20 30 40 50 60
2 ()
I
n
t
e
n
s
i
t
y

(
a
r
b
i
t
r
a
r
y

u
n
i
t
s
)
(012)
(202)
(024)
(a)
(b)
(c)
(110)
1 1
1 1
and the effect of a dynamic oxide removal process carried out within the environ-
mental chamber of the TEM.
95
The amorphous bismuthoxide layer coating of the
nanowire (Fig. 3.15(a)) is removed by exposure of the wire to hydrogen gas within the
environmental chamber of the TEM, as indicated in Fig. 3.15(b).
By coupling the powerful imaging capabilities of TEM with other characteriza-
tion tools, such as an electron energy loss spectrometer (EELS) or an energy dis-
persive x-ray spectrometer within the TEM instrument, additional properties of the
nanowires can be probed with high spatial resolution. With the EELS technique, the
energy and momentum of the incident and scattered electrons are measured in an
inelastic electron scattering process to provide information on the energy and momen-
tum of the excitations in the nanowire sample. Figure 3.16 shows the dependence
on nanowire diameter of the electron energy loss spectra of Bi nanowires.
96
The
spectra were taken from the center of the nanowire, and the shift in the energy of
the peak position (Fig. 3.16) is used to monitor the plasmon frequency in the bismuth
nanowires as the nanowire diameter is varied. The results show changes in the
electronic structures of Bi nanowires as the wire diameter decreases. Such changes
in electronic structure as a function of nanowire diameter are related to quantum
confinement effects that are also observed in their transport and optical properties,
as discussed in the next section.
3.3.3 ELECTRONIC CHARACTERIZATION
Bismuth in bulk form is a semimetal, and therefore when undoped, bulk bismuth
has an equal number of electrons and holes, each with highly anisotropic constant
energy surfaces. The corresponding highly anisotropic Fermi surfaces make bismuth
FIGURE 3.15 High-resolution transmission electron microscope image of a Bi nanowire (a)
before and (b) after annealing in hydrogen gas at 130qC for 6 h within the environmental
chamber of the instrument to remove the oxide surface layer.
95
Oxide
layer
Before After H
2
annealing at 130C
for 6 hours
(a) (b)
very interesting from an applications standpoint, because this aspect of bismuth
makes the electronic properties of bismuth nanowires so different from bulk bismuth.
Because of the very small effective mass components of bismuth, the subbands
formed under quantum confined conditions are widely spaced in energy. Thus, as
the nanowire diameter decreases, the lowest conduction subband increases in energy,
while the highest valence subband decreases in energy, thereby decreasing the band
overlap between the valence and conduction band, which is 38 meV for bulk bismuth
and for Bi nanowires of very large diameter (Fig. 3.17). The figure further shows
that at a nanowire diameter of about 50 nm, the band overlap at 77 K goes to zero
and bismuth nanowires become semiconducting. As the nanowire diameter further
decreases, the bandgap of the semiconducting nanowires increases, as is also seen
in Fig. 3.17. The figure also shows that the diameter where the semimetalsemicon-
ductor transition occurs is dependent on the crystalline orientation of the nanowire
along its axis, with the transition occurring at the largest diameter (55 nm) for wires
oriented along the highest symmetry direction, which is the three-fold trigonal
symmetry direction.
What is exceptional about bismuth is that because of the highly anisotropic
constant energy surfaces, the effective mass components normal to the light mass
component can be quite heavy. This means that it is possible to achieve very high
carrier mobility along the nanowire axis by having conduction along a high mobility,
low effective-mass crystallographic direction, while at the same time having a
reasonably high density of states, associated with the heavy carrier masses. Such an
unusual situation is very beneficial for applications such as thermoelectricity.
It would seem that the best way to study these anisotropic properties would be
through measurements of the temperature dependence of the resistivity of individual
bismuth nanowires. However, the incorporation of bismuth nanowires into test device
FIGURE 3.16 Electron energy loss spectra taken from the centers of bismuth nanowires with
diameters of 35, 60, and 90 nm. The shift in the volume plasmon peaks is due to the wire
diameter effects on the electronic structure.
96
8 10 12 14 16 18 20 22
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
Energy loss (eV)
35 nm
60 nm
90 nm
structures for such measurements has not yet been explored extensively. Two- and
four-point resistance test structures have been prepared (see Fig. 3.18) by photoli-
thography (or e-beam lithography), and metal deposition of contacts has been
achieved on individual bismuth nanowires dispersed on an oxidized silicon wafer
substrate.
97,98
The challenge in studying the temperature dependence of individual
bismuth nanowires has been the fabrication of reliable contacts to these nanowires,
because of the native oxide coating that forms on their surfaces. Progress thus far
has been made in developing reliable ohmic contacts that are stable over the 2 to
400 K temperature range. A method for avoiding the burnout of bismuth nanowires
due to electrostatic discharge has also been developed.
97
The experimental verication of the transition from semimetallic bulk bismuth
to a semiconducting nanowire can be observed in transport measurements of the
temperature-dependent resistance of nanowire arrays, as shown in Fig. 3.19(a).
Here the temperature-dependent resistance is normalized to its value at 300 K
because the number of nanowires in the template to which electrical contact is
made is not known. Figure 3.19(a) shows that the temperature-dependent normalized
resistance R(T)/R(300 K) changes radically from the curve for the bulk sample to
that for a 200-nm diameter nanowire sample, where a peak in the normalized
FIGURE 3.17 Calculated bandgap energy between the lowest electron subband and the
highest hole subband of Bi nanowires oriented along the (a) trigonal, (b) [01 2], (c) [01 1],
(d) bisectrix, and (e) binary crystallographic directions as functions of wire diameter. The
nonparabolic effects of the L-point electron pockets have been taken into account, and
cylindrical wire boundary conditions were used in the calculations. The values of d
c
, the
critical wire diameter, where the semimetal-to-semiconductor transition occurs, are 55.1, 50.0,
49.0, 48.5, and 39.8 nm for nanowires oriented along the trigonal, [01 2], [01 1], bisectrix,
and binary directions, respectively. The curves for the bisectrix and [01 1] directions are too
close to one another to be resolved.
94
0 50 100 150
Wire diameter (nm)
40
20
0
20
40
60
80
B
a
n
d

g
a
p

E
g

(
m
e
V
)
(a) Trigonal
(b) [0112]
(c) [1011]
(d) Bisectrix
(e) Binary
38
1 1
1 1
1
R(T)/R(300 K) curve appears. The peak gains prominence as fewer and fewer subbands
contribute to the carrier conduction, as shown by the experimental results for the
70-nm diameter nanowire. This behavior is further confirmed by a model calculation
of the normalized resistance R(T)/R(300 K) (Fig. 3.19(b)). Finally when the wire diam-
eter enters into the semiconductor regime below 50 nm (Fig. 3.17), the experimental
FIGURE 3.18 SEM image of four contacts to a single bismuth nanowire used for transport
measurements.
98
FIGURE 3.19 (a) Measured T dependence of the normalized resistance R(T)/R(300 K) for
highly crystalline Bi nanowires of various d
w
.
92
(b) Calculated R(T)/R(300 K) for 36- and
70-nm Bi nanowires.
78
The dashed line refers to a 70-nm polycrystalline wire with increased
grain boundary scattering.
T (K)
10 0 100
T (K)
10 100
0
1
2
3
4
R
(
T
)
/
R
(
2
7
0

K
)
36 nm
70 nm
70 nm
2
0
0

n
m
28 nm
35 nm
48 nm
Wire diameter
70 nm (polycrystalline)
(b) (a)
0.0
0.5
1.0
1.5
2.0
2.5
R
(
T
)
/
R
(
T

=

3
0
0

K
)
Bulk
trigonal Bi
curves for R(T)/R(300 K) in Fig. 3.19(a) show a different monotonic functional form,
as seen by comparing the experimental curves in this lower diameter range with
results shown for wire diameters of 200 nm and 70 nm.
92
Model calculations for
nanowires in the semiconducting regime (Fig. 3.19(b)) reproduce the observed form
of R(T)/R(300 K).
78
To further verify the validity of the model calculations, a curve
for R(T)/R(300 K) for polycrystalline bismuth wires was calculated (Fig. 3.19(b))
and these results were also in good agreement with measurements of Hong et al.
99
Further evidence for the semimetalsemiconductor transition in bismuth nano-
wires comes from measurements made on randomly oriented bismuth nanowires
of very small diameter prepared in the pores of silica, Vycor glass, and alumina
templates.
72
The wire diameters in this case were sufciently small so that the
bandgaps were large enough to allow large resistance changes of several orders of
magnitude over the measured temperature range, so that a tting of the experimental
points could be made to the functional form R(T ) = R
0
exp(E
g
/kT ), as shown in
Fig. 3.20(a), in order to obtain a measure of the bandgap. The results thus obtained
for the semiconducting bandgap are shown in Fig. 3.20(b),
72
and good agreement
was obtained for these experimental results with theoretical predictions
100
for the
FIGURE 3.20 (a) Temperature dependence of the normalized resistance R(T)/R(300 K) for
randomly oriented Bi nanowires with various diameters in the range d
w
15 nm in silica,
alumina, and Vycor glass templates, as indicated in the gure. The points are experimental data,
while the lines are ts to R(T) = R
0
exp(E
g
/kT).
72
(b) Calculated
100
values for the energy gaps
of Bi nanowires as functions of wire diameter d
w
for two different crystallographic orientations
of the nanowire axis (Fig. 3.17). The points are the experimental values of E
g
obtained from
ts to R(T) = R
0
exp (E
g
/kT) tted to three different samples as shown in (a).
72
1000
500
100
10
1
0.1
0
5 10 15 0 100 200 300
E
g

(
m
e
V
)
1000
100
10
1
R
(
T
)
/
R
(
3
0
0

K
)
T (K) d
w
(nm)
Theory,
trigonal direction
Theory,
[0-112]
Exp. transport
(a) (b)
E
2
= 0.39 eV
15 nm, SiO
2
4 nm, Bi/Vycor
E
2

=

0
.
1
9

e
V
9 nm,
Al
2
O
2
temperature dependence of the semiconducting bandgap. Using these very thin
bismuth nanowires allowed exploration for carrier localization effects at low
temperature.
75
The semimetalsemiconductor transition in nanowires can also be probed by
examination of the effect of doping. In the case of bismuth nanowires, doping with
isoelectronic antimony allows variation of the nanowire diameter at which the
semimetal-to-semiconductor phase transition occurs. As a 3-D system, Bi
1x
Sb
x
alloys form a substitutional 3-D solid-state system over the entire range of compo-
sitions x. As shown in Fig. 3.21(a) the band overlap of bismuth decreases upon Sb
addition and the alloys become semiconductors for x in the range 0.07 x 0.22
with a bandgap that depends on x. For 0.09 x 0.16, Fig. 3.21(a) shows that the
Bi
1x
Sb
x
bulk alloy is a direct gap semiconductor, with both electron and hole pockets
located at the L points in the Brillouin zone (see inset of Fig. 3.21(b)). In the range
0.07 x 0.09, the Bi
1x
Sb
x
bulk alloy is an indirect gap semiconductor with a hole
pocket at the T point, while in the range 0.16 x 0.22, the alloy is an indirect
gap semiconductor with holes forming around the six H-points shown in the inset
of Fig. 3.21(b). Therefore by antimony addition, the electronic band structure of
Bi
1x
Sb
x
can be modified over a wide range of electronic behaviors, suggesting that
Bi
1x
Sb
x
alloy nanowires should constitute an interesting 1-D system in which the
electronic band structure and related properties can be varied over an even wider
range of behaviors than for 3-D Bi
1x
Sb
x
by combining the quantum confinement
effect with the Sb alloying effect.
Figure 3.21(b) displays the phase diagram of Bi
1x
Sb
x
nanowires as a function of
wire diameter and Sb concentration.
101
This phase diagram is calculated
101
assuming
the same unit cell geometry as occurs in the parent bulk Bi
1x
Sb
x
alloy, consistent
with x-ray diffraction measurements. As can be seen from Fig. 3.21(b), there are two
regions where a semimetal-to-semiconductor phase transition occurs, one involving
L-point electrons and T-point holes (upper left of Fig. 3.21(b)), out to nearly x 0.05
for a 100-nm nanowire, and the second region involving L-point electrons and H-
point holes (upper right of Fig. 3.21(b)), extending to x 0.24 for a 100-nm nanowire,
but providing a semiconducting phase for small-diameter wires over a much larger
Sb concentration range than in the bulk alloy.
101
The inset in Fig. 3.21(b) shows the
Brillouin zone of the Bi
1x
Sb
x
alloy system and the locations of the L, H, and T points
in the rhombohedral Brillouin zone. The semimetal-to-semiconductor phase transition
involving the T-point holes and the L-point electrons, as indicated by the phase
boundary line in Fig. 3.21(b), have been probed experimentally by temperature-
dependent normalized resistance measurements, as discussed subsequently. The crit-
ical wire diameter d
c
for the semimetal-to-semiconductor transition is about 50 nm
for pure bismuth nanowires at 77 K, but d
c
increases with increasing Sb concentration
for Bi
1x
Sb
x
alloy nanowires. We note from Fig. 3.21(b) that the x range of the
semiconducting phase for Bi
1x
Sb
x
nanowires broadens as the wire diameter decreases.
The semimetal-to-semiconductor phase transition was determined in the Bi
1x
Sb
x
nanowire system by measurement of the normalized temperature-dependent resis-
tance R( T )/ R(270 K) for 65-nm diameter nanowires of different Sb alloy compo-
sition x, as shown in Fig. 3.22. In this figure the measured normalized resistance of
pure bismuth shows a non-monotonic T dependence with a maximum at ~70 K,
while the resistivities of the alloy nanowire samples (with 5 at.% and 10 at.% Sb)
decrease monotonically with increasing T, characteristic of the semiconducting
phase (Fig. 3.19(a)). Furthermore, R(T)/ R(270 K) for the Bi
0.95
Sb
0.05
nanowires
displays a stronger T dependence than that of the Bi
0.90
Sb
0.10
nanowires. The satu-
ration of R(T)/ R(270 K) at low T for all the three nanowire compositions is
attributed to carriers associated with uncontrolled impurities that are estimated to
FIGURE 3.21 Diagrams for the electronic phases for (a) 3-D bulk Bi
1x
Sb
x
plotted as energy
versus composition
75,102
and for (b) Bi
1x
Sb
x
nanowires as a function of wire diameter and Sb
concentration.
101
L
a
L
a
L
a
L
s
L
s
L
s
1
3
.
6

m
e
V
3
8

m
e
V
E
Bismuth Antimony
T
0.04
0.07
0.09
T
H
x
H
0.150.17
0.22
2
0
0

m
e
V
1
7
7

m
e
V
SC SM SM
SM: semimetal
SC: semiconductor
3
6

m
e
V
(a)
20
10
0 5 10 15 20 25 30
30
40
50
60
70
80
90
Indirect
semiconductor
Indirect
semiconductor
Direct
semiconductor
100
W
i
r
e

d
i
a
m
e
t
e
r

(
n
m
)
Semimetal Semimetal
Antimony content (at %)
H
T
L
(b)
have a concentration of about 1 to 4 u 10
16
/cm
3
. The carriers associated with the
uncontrolled impurities become increasingly important as the bandgap of the semi-
conductor increases (Fig. 3.17). Based on the calculated temperature-dependent
carrier concentration and the measured normalized temperature-dependent resis-
tance R( T ) / R(270 K) shown in Fig. 3.22, the temperature-dependent average
mobility ratio P( T ) / P(100 K) was obtained for the three samples. These calcula-
tions were then used to predict the Sb concentration at which the semimetal to
semiconductor transition occurs, as indicated in Fig. 3.22, and the result was found
to be in very good agreement with the theoretical prediction for this transition at
x 0.02 for a 65-nm Bi
1x
Sb
x
nanowire, as shown in Fig. 3.21(b). Confirmation
of the phase diagram shown in Fig. 3.21(b) allows the diagram to be used for
guidance in designing nanowires in the semiconducting phase with some control
of the bandgap of semiconducting bismuth as well as some control of other nanowire
properties.
Measurements of the temperature-dependent Seebeck effect using the smallest
available thermocouples for temperature sensors, as shown in Fig. 3.23(a), provide
further support for the phase diagram shown in Fig. 3.21(b). The data in Fig. 3.23(b)
show an increase in the magnitude of the Seebeck coefficient either as the wire diameter
of bismuth nanowires decreases or as Sb is added to form a Bi
1x
Sb
x
alloy. Further
enhancement of the magnitude of the Seebeck coefficient is achieved when both the
nanowire diameter decreases and Sb is added, as shown in Fig. 3.23(b).
103,104
FIGURE 3.22 Measured temperature dependence of the 270 K normalized resistance of
65-nm Bi
1x
Sb
x
nanowires with different Sb contents. The inset shows ratios of the mobility
P(10 K)/P(100 K) between 10 K and 100 K for the three samples (x) and the ratios of the
resistance R(10 K)/R(100 K) () as a function of Sb content, where the points are exper-
imental values and the curves are predictions.
103
1 10 100
Temperature (K)
0.0
0.5
1.0
1.5
2.0
2.5
R
(
T
)
/
R
(
2
7
0

K
)
0.00 0.05 0.10
x
0
2
4
(
1
0

K
)
/
(
1
0
0

K
)
0
2
4
R
(
1
0

K
)
/
R
(
1
0
0

K
)
0% Sb
10% Sb
5% Sb
3.3.4 OPTICAL CHARACTERIZATION
Even though the bismuth nanowire diameter is typically an order of magnitude smaller
than the wavelength of light, optical reflectivity and transmission measurements have
nevertheless been used successfully to characterize bismuth nanowires.
105,106
What
is surprising is that the intense and sharp absorption peak that dominates the infrared
absorption spectrum in bismuth nanowires,
105,106
is not observed in bulk bismuth.
The energy position E
p
of this strong absorption peak increases with decreasing
diameter. However, the rate of increase in energy with decreasing diameter |wE
p
/ wd
W
|
is an order of magnitude less than that predicted for either a direct interband
transition or for intersubband transitions in bismuth nanowires. On the other hand,
the magnitude of |wE
p
/ wd
W
| agrees well with that predicted for an indirect L-point
valence to T-point valence band transition (see inset to Fig. 3.21(b)). Since both
the initial and final states for the indirect L-T point valence band transition downshift
in energy as the wire diameter d
W
is decreased, the shift in the absorption peak
results from a difference between the effective masses for these two carrier pockets
and is less sensitive to the actual value of either of the masses. Hence the diameter
dependence of the absorption peak energy is an order of magnitude less for a valence-
to-valence band indirect transition than for a direct interband L-point transition.
Furthermore the band-tracking effect for this indirect transition gives rise to a large
value for the joint density of states, thus accounting for the high intensity of this
feature. This indirect transition arises through the gradient of the dielectric function,
which is large at the bismuthair or bismuthalumina interfaces. In addition, the
intensity can be quite significant because abundant initial state electrons, final state
holes, and appropriate phonons are available for making an indirect L-T point valence
band transition at room temperature. Interestingly the polarization dependence of
this absorption peak is such that the strong absorption is present when the electric
field is perpendicular to the wire axis, but is absent when the electric field is parallel
FIGURE 3.23 (a) Experimental setup of Seebeck coefficient measurement for nanowire
arrays. (b) Measured Seebeck coefficient as a function of T for Bi and Bi
0.95
Sb
0.05
nanowires
with different diameters. The solid curve denotes the Seebeck coefficient for bulk Bi.
103
0 100 200 300
Temperature (K)
80
60
40
20
0
S

(
V
/
K
)
65-nm Bi
40-nm Bi
45-nm Bi
0.95
Sb
0.05
65-nm Bi
0.95
Sb
0.05
Bulk Bi
Heat sink
To voltmeter
Heater
Nanowire
sample
Thermocouple
(b) (a)
To voltmeter
to the wire axis, contrary to a traditional polarizer, such as a carbon nanotube where
the optical E field is polarized by the nanotube itself to be aligned along the carbon
nanotube axis. The observed polarization dependence for bismuth nanowires is
consistent with a surface-induced effect that increases the coupling between the L-
point and T-point bands throughout the full volume of the nanowire. The indirect L
to T point valence band transition mechanism
105,106
is also consistent with observa-
tions of the effect on the optical spectra of a decrease in the nanowire diameter and
of n-type doping of bismuth nanowires with tellurium.
105
3.3.5 FUTURE DIRECTIONS FOR BISMUTH NANOWIRES
Bismuth nanowires are likely to provide a model system for future nanowire studies
because of their unique properties, namely their highly anisotropic constant energy
surfaces that allow these nanowires to simultaneously have high mobility carriers at
a reasonable high electronic density of states and to exhibit quantum effects at
relatively high nanowire diameter. The anisotropy associated with the nanowire
geometric structure breaks the symmetry so that the wire axis can be chosen to be
along a low effective masshigh carrier mobility direction simultaneously. In this
way, the highly anisotropic constant energy contours of bismuth can be explicitly
exploited to yield the quantum confinement of the carriers that is responsible for
the various semimetal-to-semiconductor electronic phase transitions. This phase
transition allows the special electronic properties to be utilized in a semiconducting
form of pure bismuth and over an extended range of wire diameter and of chemical
composition in Bi
1x
Sb
x
alloy nanowires. Using the same dopants used in bulk
bismuth to add electrons (from column IV in the periodic table) and to add holes
(from column VI), semiconducting devices based on this nanowire system should
be possible, for example, for thermoelectric applications.
Future progress in this field will be significantly enhanced by developing better
bismuth nanowire synthesis methods yielding nanowires that are highly crystalline
(so that the anisotropic electronic energy band structure can be exploited) and to
which good electrical contacts can be made (circumventing the problems caused by
surface oxidation when Bi nanowires are exposed to ambient conditions). If tem-
plates are used for nanowire self-assembly, electrical isolation and crystallographic
alignment of nanowire arrays, more rugged template materials are needed, and for
thermoelectric applications, template materials with lower thermal conductivity
would be needed. The ability to control material properties that are linked together
in bulk bismuth but are not so closely linked in bismuth nanowires requires further
exploration through model-based selection of wire diameter and Bi
1x
Sb
x
alloy com-
position.
3.3.6 CONCLUDING REMARKS
When considering the one-dimensionality of bismuth, some comment is in order about
the ability of bismuth (and many other materials systems for that matter) to form
nanotubes and nanowires by a variety of synthesis approaches, including growth in
templates initiated at the pore interfaces and using very short growth times.
107
One
interesting aspect of a bismuth nanotube is that the wall thickness can be a single
basic structural unit consisting of two atomic layers. The diffraction pattern for
bismuth nanotubes reveals lattice constants similar to those of 3-D bismuth. Research
on the physical properties of bismuth nanotubes is at an early stage, and the electronic
structure remains to be explored. Study of the electronic structures of bismuth
nanotubes in relation to those of bismuth nanowires with similar local atomic
arrangements and similar outer diameters should prove to be fascinating. Also
intriguing should be the comparison of the electronic structure of a bismuth nanotube
to that of a carbon nanotube. Although bismuth and graphite are both semimetals
as 3-D crystals, their electronic structures as 1-D nanowires and nanotubes, respec-
tively, are very different, and it is expected that the electronic structure of a bismuth
nanotube will differ significantly from its nanowire counterpart, as well as from that
of a carbon nanotube.
From these considerations we can conclude that we are now within reach of
studying the rich variety of electronic structures that 1-D systems might exhibit,
learning how to realize this variety of structures experimentally and how to exploit
these different electronic structures to yield interesting materials properties not yet
achieved in any system, and someday be able to utilize these special properties for
practical applications.
ACKNOWLEDGMENTS
A.J. acknowledges financial support by PRPq-UFMG, and the Instituto de Nano-
cincias (Millennium Institute Program), CNPq, Brazil. O.R. and M.S.D. acknowl-
edge support under NSF Grant DMR 04-05538.
PROBLEMS
1. Based on the cutting lines concept, explain why for an (n, m) single-wall
carbon nanotube (SWNT) the nanotube is a large bandgap semiconductor,
when n m is not a multiple of three, with the band gap increasing as
the diameter decreases (1/d
t
dependence), while the nanotube shows
metallic behavior (at room temperature) for n m 3j ( j integer).
Compare the electronic density of states near the Fermi level for semi-
conducting and metallic SWNTs and for a 2-D graphene sheet.
2. How are single-wall carbon nanotubes formed in the laboratory? Which
mechanism for SWNT growth (catalytic root growth or tip growth) is
most probable for the synthesis of very long (~cm) tubes? What are the
major challenges that need to be addressed in advancing carbon nanotube
synthesis?
3. List the common methods for nanotube characterization, and their advan-
tages and limitations. What is it that each characterization method probes
and why is it important to gain this information?
4. Why is Raman spectroscopy a particularly valuable tool for the study and
characterization of carbon nanotube systems? What are its advantages and
limitations?
5. Using the tight binding model, find an expression for the electronic energy
levels of an armchair nanotube. Make an explicit application of your
expression to a (10,10) nanotube.
6. For very small-diameter nanotubes, at what diameter does the tube cur-
vature impose sp
3
bonding rather than sp
2
bonding. Explain in what sense
we can consider this cross-over point to indicate that the graphitic nano-
tube becomes a nanodiamond.
7. Find the electronic energy levels for a free electron nanowire with a
square cross section (a
2
) and a length 10a. At what temperature would
you expect quantum confinement effects to become important at room
temperature?
8. Find the electronic energy levels of a silicon nanowire of diameter d
w
with its wire axis along (100) and then repeat for its wire axis along
(111). Use the effective mass approximation and effective mass tensor
components for 3-D Si. Explain how to find the effective mass that
governs transport along the nanowire axis and how to find the threshold
for optical transitions.
9. Assume that the carriers of a nanowire of diameter d
w
have an effective
mass of 0.1m
0
in the plane of the nanowire and 0.05m
0
along the nanowire
axis. At what magnetic field value would the electron scattering rate for
carriers in a nanowire of diameter d
w
decrease rapidly for a magnetic field
parallel to the nanowire axis? Repeat for the case of a magnetic field
perpendicular to the wire axis.
10. Consider a silicon nanowire to be the same as in Problem 7. At what
magnetic field along the nanowire axis is the spacing of the Landau levels
(magnetic energy levels) equal to the subband spacing? Explain why
carrier localization is expected as the nanowire cross-sectional area
becomes very small.
REFERENCES
1. Dresselhaus, M.S., Dresselhaus, G., and Eklund, P.C., Science of Fullerenes and
Carbon Nanotubes, Academic Press, New York, 1996.
2. Dresselhaus, M.S., Dresselhaus, G., and Avouris, Ph., Carbon Nanotubes: Synthesis,
Structure, Properties and Applications, Vol. 80 of Springer Series in Topics in Applied
Physics, Springer-Verlag, Berlin, 2001.
3. Harris, P., Carbon Nanotubes and Related Structures: New Materials for the Twenty-
First Century, Cambridge University Press, Cambridge, UK, 2001.
4. Bandow, S., Hiraoka, K., Chen, G., Eklund, P.C., and Iijima, S., Bull. Mater. Res.
Soc., 29, 260264, 2004.
5. Saito, R., Dresselhaus, G., and Dresselhaus, M.S., Physical Properties of Carbon
Nanotubes, Imperial College Press, London, 1998.
6. Dresselhaus, M.S., Dresselhaus, G., Charlier, J.C., and Hernndez, E., Phil. Trans. Roy.
Soc. A, 362, 20652098, 2003.
7. Hamada, N., Sawada, S., and Oshiyama, A., Phys. Rev. Lett., 68, 15791581, 1992.
8. Saito, R., Fujita, M., Dresselhaus, G., and Dresselhaus, M.S., Appl. Phys. Lett., 60,
22042206, 1992.
9. Mintmire, J.W., Dunlap, B.I., and White, C.T., Phys. Rev. Lett., 68, 631634, 1992.
10. Samsonidze, Ge.G., Grneis, A., Saito, R., Jorio, A., Souza Filho, A.G., Dresselhaus,
G., and Dresselhaus, M.S., Phys. Rev. B, 69, 205402, 111, 2004.
11. Samsonidze, Ge.G., Saito, R., Jorio, A., Pimenta, M.A., Souza Filho, A.G., Grneis,
A., Dresselhaus, G., and Dresselhaus, M.S., J. Nanosci. Nanotechnol., 3, 431458,
2003.
12. Chico, L., Crespi, V.H., Benedict, L.X., Louie, S.G., and Cohen, M.L., Phys. Rev.
Lett., 76, 971974, 1996.
13. Liu, J., Fan, S., and Dai, H., Bull. Mater. Res. Soc., 29, 244250, 2004.
14. Bethune, D.S., Kiang, C.H., de Vries, M.S., Gorman, G., Savoy, R., Vazquez, J., and
Beyers, R., Nature (London), 363, 605, 1993.
15. Journet, C., Maser, W.K., Bernier, P., Loiseau, A., Lamy de la Chapelle, M., Lefrant,
S., Deniard, P., Lee, R., and Fischer, J.E., Nature (London), 388, 756758, 1997.
16. Thess, A., Lee, R., Nikolaev, P., Dai, H., Petit, P., Robert, J., Xu, C., Lee, Y.H., Kim,
S.G., Rinzler, A.G., Colbert, D.T., Scuseria, G.E., Tomnek, D., Fischer, J.E., and
Smalley, R.E., Science, 273, 483487, 1996.
17. Nilolaev, A.V., Bronikowski, M.J., Bradley, R.K., Rohmund, F., Colbert, D.T., Smith,
K.A., and Smalley, R.E., Chem. Phys. Lett., 313, 9197, 1999.
18. Maruyama, S., Physica B, 323, 193195, 2002.
19. Hayashi, T., Kim, Y.A., Matoba, T., Esaka, M., Nishimura, K., Endo, M., and
Dresselhaus, M.S., Nano Lett., 3, 887889, 2003.
20. Kitiyanan, B., Alvarez, W.E., Harwell, J.H., and Resasco, D.E., Chem. Phys. Lett.,
317, 497, 2000.
21. Alvarez, W.E., Kitiyanan, B., Borgna, A., and Resasco, D.E., Carbon, 39, 547, 2001.
22. Haddon, R.C., Sippel, J., Rinzler, A.G., and Papadimitrakopoulos, F., Bull. Mater.
Res. Soc., 29, 252259, 2004.
23. Liu, K., Chien, C.L., and Searson, P.C., Phys. Rev. B, 58, R14681R14684, 1998.
24. Rinzler, A.G., Liu, J., Dai, H., Nikolaev, P., Huffman, C.B., and Rodriguez-Marcias,
F.J., Appl. Phys. A, 67, 2937, 1998.
25. Chattopadhyay, D., Galeska, I., and Papadimitrakopoulos, F., Carbon, 40, 985988,
2002.
26. Endo, M., Kim, Y.A., Fukai, Y., Hayashi, T., Terrones, M., Terrones, H., and Dresselhaus,
M.S., Appl. Phys. Lett., 79, 15311533, 2001.
27. Endo, M., Hayashi, T., Kim, Y.A., Terrones, M., and Dresselhaus, M.S., Phil. Trans.
Roy. Soc. 362, 22232238, 2004
28. Chattopadhyay, D., Galeska, I., and Papadimitrakopoulos, F., J. Am. Chem. Soc., 125,
33703375, 2003.
29. Zheng, M., Jagota, A., Semke, E.D., Diner, B.A., McLean, R.S., Lustig, S.R.,
Richardson, R.E., and Tassi, N.G., Nature Mater., 2, 338342, 2003.
30. Zheng, M., Jagota, A., Strano, M.S., Barone, P., Chou, S.G., Diner, B.A., Dresselhaus,
M.S., McLean, R.S., Onoa, G.B., Santos, A.P., Semke, E.D., Usrey, M., and Walls,
D.J., Science, 302, 15451548, 2003.
31. Lebedkin, S., Arnold, K., Hennrich, F., Krupke, R., Renker, B., and Kappes, M.M.,
New J. Phys., 5, 140, 2003.
32. Strano, M.S., J. Am. Chem. Soc., 125, 1614816153, 2003.
33. Strano, M.S., Miller, M.K., Allen, M.J., Moore, V.C., O'Connell, M.J., Kittrell, C.,
Hauge, R.H., and Smalley, R.E., J. Nanosci. Nanotechnol., 3, 8186, 2003.
34. Strano, M.S., J. Am. Chem. Soc., 125, 1614816150, 2004.
35. Chen, G., Sumanasekera, G.U., Pradhan, B.K., Gupta, R., Eklund, P.C., Bronikowski,
M.J., and Smalley, R.E., J. Nanosci. Nanotech., 2, 621626, 2002.
36. Kataura, H., Kimura, A., Ohtsuka, Y., Suzuki, S., Maniwa, Y., Hanyu, T., and Achiba,
Y., Jpn. J. Appl. Phys., 37, L616L618, 1998.
37. Hagen, A., and Hertel, T., Nanoletters, 3, 383388, 2003.
38. Rao, A.M., Richter, E., Bandow, S., Chase, B., Eklund, P.C., Williams, K.W., Fang,
S., Subbaswamy, K.R., Menon, M., Thess, A., Smalley, R.E., Dresselhaus, G., and
Dresselhaus, M.S., Science, 275, 187191, 1997.
39. Dresselhaus, M.S., Dresselhaus, G., Jorio, A., Souza Filho, A.G., and Saito, R.,
Carbon, 40, 20432061, 2002.
40. O'Connell, M.J., Bachilo, S.M., Huffman, X.B., Moore, V.C., Strano, M.S., Haroz,
E.H., Rialon, K.L., Boul, P.J., Noon, W.H., Kittrell, C., Ma, J., Hauge, R.H., Weisman,
R.B., and Smalley, R.E., Science, 297, 593596, 2002.
41. Jorio, A., Saito, R., Hafner, J.H., Lieber, C.M., Hunter, M., McClure, T., Dresselhaus,
G., and Dresselhaus, M.S., Phys. Rev. Lett., 86, 11181121, 2001.
42. Dresselhaus, M.S., and Eklund, P.C., Advan. Phys., 49, 705814, 2000.
43. Pimenta, M.A., Marucci, A., Empedocles, S., Bawendi, M., Hanlon, E.B., Rao, A.M.,
Eklund, P.C., Smalley, R.E., Dresselhaus, G., and Dresselhaus, M.S., Phys. Rev. B
Rapid, 58, R16016R16019, 1998.
44. Brown, S.D.M., Jorio, A., Corio, P., Dresselhaus, M.S., Dresselhaus, G., Saito, R.,
and Kneipp, K., Phys. Rev. B, 63, 155414, 2001.
45. Jorio, A., Dresselhaus, G., Dresselhaus, M.S., Souza, M., Dantas, M.S.S., Pimenta,
M.A., Rao, A.M., Saito, R., Liu, C., and Cheng, H.M., Phys. Rev. Lett., 85,
26172620, 2000.
46. Jorio, A., Pimenta, M.A., Souza Filho, A.G., Samsonidze, Ge.G., Swan, A.K., nl,
M.S., Goldberg, B.B., Saito, R., Dresselhaus, G., and Dresselhaus, M.S., Phys. Rev.
Lett., 90, 107403, 2003.
47. Thomsen, C., and Reich, S., Phys. Rev. Lett., 85, 5214, 2000.
48. Souza Filho, A.G., Jorio, A., Samsonidze, Ge.G., Dresselhaus, G., Dresselhaus, M.S.,
Swan, A.K., nl, M.S., Goldberg, B.B., Saito, R., Hafner, J.H., Lieber, C.M., and
Pimenta, M.A., Chem. Phys. Lett., 354, 6268, 2002.
49. Saito, R., Dresselhaus, G., and Dresselhaus, M.S., Phys. Rev., B, 61, 29812990, 2000.
50. Samsonidze, Ge.G., Saito, R., Jorio, A., Souza Filho, A.G., Grneis, A., Pimenta,
M.A., Dresselhaus, G., and Dresselhaus, M.S., Phys. Rev. Lett., 90, 027403, 2003.
51. Kataura, H., Kumazawa, Y., Maniwa, Y., Umezu, I., Suzuki, S., Ohtsuka, Y., and
Achiba, Y., Synthet. Met., 103, 25552558, 1999.
52. Souza Filho, A.G., Chou, S.G., Samsonidze, Ge.G., Dresselhaus, G., Dresselhaus,
M.S., Lei An, J. Liu, Anna Swan, K., nl, M.S., Goldberg, B.B., Jorio, A., Grneis,
A., and Saito, R., Phys. Rev. B, 69, 115428, 2004.
53. Jorio, A. , Souza Filho, A.G., Dresselhaus, G., Dresselhaus, M.S., Saito, R., Hafner,
J.H., Lieber, C.M., Matinaga, F.M., Dantas, M.S.S., and Pimenta, M.A., Phys. Rev.
B, 63, 245416, 2001.
54. Fantini, C., Jorio, A., Souza, M., Mai Jr., A.J., Strano, M.S., Dresselhaus, M.S., and
Pimenta, M.A., Phys. Rev. Lett., 93, 147406, 2004.
55. , Tinkham, M., Chou, S.G., Rabin, O., Dresselhaus, M.S.,
Swan, A.K., nl, M.S., and Goldberg, B.B., Appl. Phys. Lett., 84, 20522055, 2004;
Virtual J. Nanoscale Sci. Technol., 2004.
56. Lin, Yu-Ming, and Dresselhaus, M.S., Appl. Phys. Lett., 83, 35673569, 2003.
57. Fantini, C., Jorio, A., Souza, M., Ladeira, L.O., Pimenta, M.A., Souza Filho, A.G.,
Saito, R., Samsonidze, Ge.G., Dresselhaus, G., and Dresselhaus, M.S., Phys. Rev.
Lett., 93, 087401, 2004.
58. Dresselhaus, G., Pimenta, M.A., Saito, R., Charlier, J.-C., Brown, S.D.M., Corio, P.,
Marucci, A., and Dresselhaus, M.S. in Tomnek, D. and Enbody, R.J., Eds., Science
and Applications of Nanotubes, 275295, Kluwer Academic, New York, 2000.
59. Hartschuh, A., Sanchez, E.J., Xie, X.S., and Novotny, L., Phys. Rev. Lett., 90, 95503,
2003.
60. Hartschuh, A., Pedrosa, H.N., Novotny, L., and Krauss, T.D., Science, 301, 1354
1356, 2003.
61. Saito, R., and Kataura, H., Carbon Nanotubes: Synthesis, Structure, Properties and
Applications, Dresselhaus, M.S., Dresselhaus, G., and Avouris, P. Eds., pp. 213246,
Springer Series in Topics in Applied Physics, Vol. 80, Springer-Verlag, Berlin, 2001.
62. Kuzmany, Hans, Solid-State Spectroscopy: An Introduction, Springer-Verlag, Berlin,
Vol. 50, p. 115. 1998.
63. Sen, R., Rickard, S.M., Itkis, M.E., and Haddon, R.C., Chem. Mater., 15, 4273, 2003.
64. Hertel, T., and Moos, G., Chem. Phys. Lett., 320, 359364, 2000.
65. Hertel, T., and Moos, G., Phys. Rev. Lett., 84, 50025005, 2000.
66. Bachilo, S.M., Strano, M.S., Kittrell, C., Hauge, R.H., Smalley, R.E., and Weisman,
R.B., Science, 298, 23612366, 2002.
67. Weisman, R.B., and Bachilo, S.M., Nanoletters, 3, 12351238, 2003.
68. Niyogi, S., Hamon, M.A., Hu, H., Zhao, B., Bhowmik, P., Sen, R., Itkis, M.E., and
Haddon, R.C., Acc. Chem. Res., 35, 11051113, 2002.
69. Feynman, Richard P. (1959). From a talk reported at http://www.zyvex.com/nanotech/
feynman.html.
70. Dresselhaus, M.S., Lin, Y.-M., Rabin, O., Black, M.R., and Dresselhaus, G., in Bhushan,
B., Ed., Springer Handbook of Nanotechnology, Springer-Verlag, Heidelberg, 2004, pp.
99145.
71. Dresselhaus, M.S., Lin, Yu-Ming, Cronin, S.B., Rabin, O., Black, M.R., Dresselhaus,
G., and Koga, T. in Semiconductors and Semimetals: Recent Trends in Thermoelectric
Materials Research III, Tritt, T. M., Academic Press, San Diego, 2001 Ed., 1 pp.
1121.
72. Heremans, J., in Thermoelectric Materials 2003Research and Applications: MRS
Symposium Proceedings, Boston, December 2003, Nolas, G.S., Yang, J., Hogan, T.P.,
and Johnson, D.C., Eds., Materials Research Society Press, Pittsburgh, 2004, pp. 314.
73. Whitney, T.M., Jiang, J.S., Searson, P.C., and Chien, C.L., Science, 261, 1316, 1993.
74. Piraux, L., Dubois, S., Duvail, J.L., and Radulescu, A., J. Mater. Res., 14, 30423050,
1999.
75. Rabin, Oded, Bismuth Nanowire and Antidot Array Studies Motivated by Thermo-
electricity, Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, June
2004.
76. Huber, C.A., Huber, T.E., Sadoqi, M., Lubin, J.A., Manalis, S., and Prater, C.B.,
Science, 263, 800802, 1994.
77. Zhang, Z., Ying, J.Y., and Dresselhaus, M.S., J. Mater. Res., 13, 17451748, 1998.
78. Lin, Y.-M., Cronin, S.B., Ying, J.Y., Dresselhaus, M.S., and Heremans, J.P., Appl.
Phys. Lett., 76, 39443946, 2000.
79. Adamson, A.W., Physical Chemistry of Surfaces, Wiley, New York, 1982, p. 338.
80. Zhang, Z., Gekhtman, D., Dresselhaus, M.S., and Ying, J.Y., Chem. Mater., 11,
16591665, 1999.
81. Huber, T.E., Graf, M.J., Foss, Jr., C.A., and Constant, P., J. Mater. Res., 15,
18161821, 2000.
82. Liu, K., Chien, C.L., Searson, P.C., and Kui, Y.Z., Appl. Phys. Lett., 73, 14361438, 1998.
83. Almawlawi, D., Liu, C.Z., and Moskovits, M., J. Mater. Res., 9, 10141018, 1994.
84. Sapp, S.A., Lakshmi, B.B., and Martin, C.R., Adv. Mater., 11, 402404, 1999.
85. Prieto, A.L., Sander, M.S., Martin-Gonzalez, M.S., Gronsky, R., Sands, T., and Stacy,
A.M., J. Am. Chem. Soc., 123, 71607161, 2001.
86. Sander, M.S., Gronsky, R., Sands, T., and Stacy, A.M., Chem. Mater., 15, 335339,
2003.
87. Yin, A.J., Li, J., Jian, W., Bennett, A.J., and Xu, J.M., Appl. Phys. Lett., 79,
10391041, 2001.
88. Li, L., Zhang, Y., Li, G., and Zhang, L., Chem. Phys. Lett. 378, 244249, 2003.
89. Jin, C.G., Jiang, G.W., Liu, W.F., Cai, W.L., Yao, L.Z., Yao, Z., and Li, X.G., J.
Mater. Chem., 13, 17431746, 2003.
90. Toimil Molares, M.E., Chtanko, N., Cornelius, T.W., Dobrev, D., Enculescu, I., Blick,
R.H., and Neumann, R., Nanotechnology, 15, S201S207, 2004.
91. Rabin, O., Chen, G., and Dresselhaus, M.S., The Electrochemical Society Meeting
Abstracts, 2003-02, Abs. 25, 2003.
92. Heremans, J., Thrush, C.M., Lin, Yu-Ming, Cronin, S., Zhang, Z., Dresselhaus, M.S.,
and Mansfield, J.F., Phys. Rev. B, 61, 29212930, 2000.
93. Rabin, O., Herz, P.R., Lin, Y.-M., Akinwande, A.I., Cronin, S.B., and Dresselhaus,
M.S., Adv. Funct. Mater., 13, 631638, 2003.
94. Yu-Ming, Lin., Thermoelectric Properties of Bi
1x
Sb
x
and Superlattice Nanowires,
Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, June 2003.
95. Cronin, S.B., Lin, Y.-M., Rabin, O., Black, M.R., Dresselhaus, G., Dresselhaus, M.S.,
and Gai, P.L., Microscopy Microanalysis, 8, 5863, 2002.
96. Sander, M.S., Gronsky, R., Lin, Y.-M., and Dresselhaus, M.S., J. Appl. Phys., 89,
27332736, 2001.
97. Cronin, S.B., Lin, Y.-M., Rabin, O., Black, M.R., Ying, J.Y., Dresselhaus, M.S., Gai,
P.L., Minet, J.-P., and Issi, J.-P., Nanotechnology J., 13, 653658, 2002.
98. Cronin, Stephen B., Electronic Properties of Bi Nanowires, Ph.D. thesis, Massachusetts
Institute of Technology, Cambridge, June 2002.
99. Hong, K., Yang, F.Y., Liu, K., Reich, D.H., Searson, P.C., and Chien, C.L., J. Appl.
Phys., 85, 61846186, 1999.
100. Lin, Y.-M., Sun, X., and Dresselhaus, M.S., Phys. Rev. B, 62, 4610 4623, 2000.
101. Rabin, O., Lin, Yu-Ming, and Dresselhaus, M.S., Appl. Phys. Lett., 79, 8183, 2001.
102. Lenoir, B., Cassart, M., Mickenand, J.P., and Scharrer, H., Scharrer, S., J. Phys. Chem.
Solids 57, 8999, 1996.
103. Lin, Y.-M., Rabin, O., Cronin, S.B., Ying, J.Y., and Dresselhaus, M.S., Appl. Phys.
Lett., 81, 24032405, 2002.
104. Lin, Y.-M., Rabin, O., Cronin, S.B., Ying, J.Y., and Dresselhaus, M.S., in Caillat,
T., and Snyder, J. 21st International Conference on Thermoelectrics: ICT Sympo-
sium Proceedings, Long Beach, CA, 2002, pp. 253256.
105. Black, Marcie R., The Optical Properties of Bi Nanowires, Ph.D. thesis, Massachu-
setts Institute of Technology, Cambridge, June 2003.
106. Black, M.R., Hagelstein, P.L., Cronin, S.B., Lin, Y.-M., and Dresselhaus, M.S., Phys.
Rev. B, 68, 235417, 2003.
107. Dresselhaus, M.S., Lin, Y.-M., Rabin, O., Jorio, A., Souza Filho, A.G., Pimenta,
M.A., Saito, R., Samsonidze, Ge.G., and Dresselhaus, G., Mater. Sci. Engin. C, 23,
129140, 2003.
4
Nanobelts
and Nanowires
of Functional Oxides
Xudong Wang and Zhong Lin Wang
CONTENTS
4.1 Introduction
4.2 The Nanobelt: What Is It?
4.3 Techniques for Growing Nanobelts/Nanowires
4.3.1 Thermal Evaporation
4.3.2 Laser AblationAssisted CVD
4.4 Growth Mechanisms
4.5 The Nanobelt Family
4.5.1 Binary Oxide Nanobelts
4.5.2 Compound Semiconductor Nanobelts
4.5.3 Multielement Nanobelts
4.6 Ultra-Narrow ZnO Nanobelts
4.7 Mesoporous ZnO Nanowires
4.8 Patterned Growth of Aligned ZnO Nanowires
4.9 Selected Applications of Nanobelts
4.9.1 Nanocantilevers
4.9.2 Field-Effect Transistors
4.9.3 Nanosensors
4.10 Summary
Problems
Acknowledgments
References
4.1 INTRODUCTION
Nanotechnology can be defined as the creation and utilization of materials, devices,
and systems through the control of matter on the nanometer (nm) scale, that is, at
the level of atoms, molecules, supermolecular structures, and with at least one
characteristic dimension measured in nanometers (1 a 100 nm). Nanomaterials and
systems are rationally designed to exhibit novel, unique, and significantly improved
physical, chemical, and biological properties, phenomena, and processes because of
their small sizes. Depending on the number of nanosized dimensions, nanomaterials
can be classified as two-dimensional (2-D) thin/nanolayers, one-dimensional (1-D)
nanowires, and zero-dimensional quantum dots. Quasi-1-D nanostructures
1
have
recently attracted considerable research interest due to their unique properties and
wide range of applications. They show great potential both for understanding size-
dependent electrical, optical, thermal, and mechanical properties and for fabricating
nanosized electrical junctions and optoelectronic and electromechanical devices. The
nanobelt,
2
an important new member in the 1-D nanomaterial family, has a well-
defined structure and is expected to be a good candidate as a 1-D nanoscale sensor,
transducer, and resonator. This chapter provides a review of current progress in
oxide-based nanowires and nanobelts, focusing on growth methods, growth mech-
anisms, properties, and potential applications. Some unique morphologies of 1-D
nanostructures, such as ultra-narrow nanobelts, mesoporous nanowires, and hexag-
onal-patterned aligned nanorods, are presented.
4.2 THE NANOBELT: WHAT IS IT?
Several terminologies are defined for 1-D nanostructures, including nanobelt, nanow-
ire, nanorod, and nanotube. These terminologies appear frequently in the literature
owing to increasing research interests in the area. The term nanowire, which means
a 1-D nanostructure that grows along a specific axial direction whose side surfaces
may not be well defined, is fairly popular. The cross section of a nanowire can be
round, hexagonal, or a polyhedron according to the crystallography of the material.
The length of a nanowire varies from a few hundred nanometers to microns or even
millimeters, and its thickness is always negligibly small compared to its length (Fig.
4.1(a)). A nanorod can be defined as a short nanowire of length in the range from
tens to hundreds of nanometers (Fig. 4.1(b)).
Since the first discovery by our group in 2001, nanobelts, sometimes called
nanoribbons, have attracted more and more research interest due to their unique
morphology and shape-induced physical, electronic, and optical properties. The nano-
belt represents a quasi-1-D nanostructure that not only has a specific growth direction
but whose top/bottom surfaces and side surfaces are also well defined crystallographic
facets (Fig. 4.1(c)). For example, a typical ZnO nanobelt
2
has a growth direction
along [0001], the top/bottom surfaces are r( ), and the side surfaces are
r( ), as shown in the of Fig. 4.1(c) insert. The as-synthesized metal oxide
nanobelts exhibit a rectangular-like cross section and are single crystalline and struc-
turally uniform. Nanobelts are nanowires that have well-defined structure. This is an
important characteristic because the properties of nanobelts are strongly shape and
structure dependent, especially when their sizes approach a few nanometers.
4.3 TECHNIQUES FOR GROWING
NANOBELTS/NANOWIRES
Synthesis of nanomaterials is one of the most active fields in nanotechnology. Various
1-D nanostructures have been fabricated using a number of approaches, including
thermal evaporation,
2
laser ablation,
3
template-assisted growth,
4,5
arc discharge,
6
21 10
0110
lithography,
7
and sol-gel method.
8
Thermal evaporation and laser ablation, which
have similar experimental setups and mechanisms, are usually the first choices of
synthesis due to their simple setups and acceptable controllability of the self-
assembly process. Both methods will be discussed in detail in this chapter.
4.3.1 THERMAL EVAPORATION
Thermal evaporation is one of the simplest, most popular, and most versatile syn-
thesis methods, and it has been very successful in fabricating nanobelts and nanow-
ires with various characteristics. The basic process of this method is sublimating
source material(s) in powder form at high temperature, and a subsequent deposition
of the vapor in a certain temperature region to form desired nanostructures.
A typical experimental system is shown in Figure 4.2. The synthesis is performed
in an alumina or quartz tube, which is located in a horizontal tube furnace. High-purity
FIGURE 4.1 Quasi-1-D nanostructures. (a) ZnO nanowires; (b) ZnO nanorods; (c) ZnO
nanobelts. The bottom right insert shows a transmission electronic microscopy image of
ZnO nanobelts and the corresponding electron diffraction pattern.
oxide powders contained in an alumina boat are loaded in the middle of the furnace,
the highest temperature region. The substrates for collecting the desired nanostruc-
tures are usually placed downstream following the carrier gas. The substrates can
be silicon wafer, polycrystalline alumina, or single crystal alumina (sapphire). Both
ends of the tube are covered by stainless steel caps and sealed with O-rings. Cooling
water flows inside the cover caps to achieve a reasonable temperature gradient in
the tube.
During the experiments, the system is first pumped down to around 10
2
torr.
Then the furnace is turned on to heat the tube to the reaction temperature at a specific
heating rate. An inert carrying gas, such as argon or nitrogen, is then introduced into
the system at a constant flow rate to bring the pressure in the tube back to 200500
torr. (Different pressures are required by different source materials and final depos-
ited nanostructures.) The reaction temperature and pressure are held for a certain
period of time to vaporize the source material and achieve a reasonable amount of
deposition. Zinc oxide, for example, can be vaporized in two different processes:
(4.1)
(4.2)
By controlling the growth kinetics it is possible to control the decomposition process.
In Reaction 4.1, the zinc vapor and oxygen are transported by the carrier gas to the
lower temperature region, and their recombination forms ZnO nanostructures. When
the reaction is completed, the furnace is turned off and the system is cooled down
to room temperature with flowing inert gas. Normally the system pressure is kept
at the reaction pressure to prevent subsequent decomposition of the grown nano-
structures.
The thermal evaporation process is basically a physical vapor deposition process
that has been successfully used for synthesizing a variety of oxide and non-oxide
nanobelts and nanowires.
9
Moreover, this system can also be used for chemical vapor
deposition (CVD) by simply applying reaction gases instead of the carrier gas and
placing substrates in the middle of the tube. For example, multiwall and single-wall
FIGURE 4.2 Thermal evaporate deposition system for synthesis of 1-D nanostructures.
Substrate
Source materials
Tube furnace
Cooling
water
Cooling
water
Carrying
gas
Pump
2 2
2
ZnO Zn O ( ) ( ) ( ) s v v +
ZnO ZnO ( ) ( ) s v
carbon nanotubes have been successfully fabricated in this system using hydrogen
and methane/acetylene as reactants.
10
A disadvantage of the thermal evaporation method is the high temperature for
controlling growth kinetics. Some modifications have been made to lower the growth
temperature. Mixing ZnO with carbon powder, which is so-called carbon-thermal
evaporation,
12
can reduce the vaporization temperature from 1300 to 800qC,
(4.3)
Metal catalysts, such as gold,
1113
tin,
14,15
and copper,
16
have also been used to grow
aligned nanostructures with reasonably controlled size.
4.3.2 LASER ABLATIONASSISTED CVD
Laser ablationassisted CVD growth has been demonstrated as an effective technique
for growing 1-D nanostructures, especially for semiconductors. Unlike the 2-D laser-
ablation deposition system, in which the substrate is normally on the top of the
target, this laser ablation system is a combination of a pulse laser and a thermal
evaporation system with a few modifications, as shown in Fig. 4.3. A quartz tube
has to be used and a fused silica window is assembled on the end cap for inputting
and aiming the laser beam. The source materials are highly pressed into a small
alumina crucible and face the laser beam at an angle of about 45q. The location of
the source materials can be either in the middle or outside of the furnace depending
on their temperature sensitivity. A focus lens with a long focal length is located in
the front of the tube window to converge the laser beam onto the target. A pulsed
laser vaporizes the source materials.
A typical experimental process is similar to that of the thermal evaporation
process. The furnace temperature is always lower and ranges from 500 to 800qC.
After the desired temperature and pressure are achieved, the laser beam is shot
onto the target and vaporizes the source materials stoichiometrically. The vapor
mixture subsequently deposits on the substrate with the help of the metal catalyst.
If the local deposition temperature is sufficiently low and the distance from the
FIGURE 4.3 Laser ablation deposition system for synthesis of 1-D nanostructures.
ZnO C Zn CO
C
( ) ( ) ( ) ( )
~
s s v v + +
800
Substrate
Tube furnace
Cooling
water
Lens
Laser
Cooling
water
Carrying
gas
Pump
substrate to the source is long, the vapor can also experience a homogeneous
nucleation and growth during the transportation process, and the final nanostructures
are collected by the substrate.
17
The laser ablation technique can be used to fabricate nanostructures with more
diverse and complex chemical composition. Due to the high power of the laser beam,
the source materials could be sublimated stoichiometrically at a relatively low
temperature. Without changing the tube or deposition temperature, the vapor density
can be easily modified by varying the power and frequency of the laser beam so
that size and growth control are achieved.
4.4 GROWTH MECHANISMS
One-dimensional nanostructures are formed by a vapor-solid or vapor-liquid-solid
(VLS) process. When there is no catalyst used in the growth, the formation of 1-D
nanostructures is dominated by a vapor-solid process, such as the growth of many
oxide nanobelts (ZnO, SnO
2
, Ga
2
O
3
, In
2
O
3
, etc.).
2,18
If a metal catalyst is introduced
into the reaction system, 1-D nanostructures will grow through a VLS process.
19
The
metal nanoparticles are liquid droplets at the growth temperature, and they act as
nucleation sites for absorbing the incoming molecular vapor. Nanobelts or nanowires
then grow out of the particles when the liquid droplets are supersaturated. Once the
growth temperature drops below the eutectic temperature of the particle and/or the
reactant vapor is no longer available, the growth is terminated.
Figure 4.4 shows the major steps of a VSL process using gold catalyst for growing
ZnO nanowires/nanobelts. A thin layer of gold or presynthesized gold nanoparticles
is used as catalyst (Fig. 4.4(a)). The gold catalyst forms liquid alloy droplets once
the deposition temperature rises above the eutectic temperature of Au and ZnO (Figure
4.4(b)). With the constant incoming ZnO vapor, the percentage of ZnO component
in the droplet increases and ultimately supersaturates. ZnO is precipitated out at the
solid-liquid interface and forms wire- or belt-like nanostructures (Fig. 4.4(c)).
Typically the 1-D nanostructures obtained through a VLS mechanism have a solid
metal ball at the growth tips. When the substrate has a crystalline surface, such as
FIGURE 4.4 Schematic steps of a VLS growth process in a Au/ZnO system.
Gold layer
Au-ZnO
alloy ball
ZnO vapor
ZnO vapor
Growth of
ZnO nanobelt
(a) (b) (c)
sapphire,
20
the as-synthesized nanowires normally have roots at the substrate and point
upward with the catalyst balls on the tips (Figure 4.5(a)), which may be a result of
epitaxial growth of ZnO on alumina. If the surface of the substrate is disordered, such
as silica or polycrystalline alumina, the catalyst balls are still at the growth fronts, but
the nanowires are randomly oriented
12
(Figure 4.5(b)). However, once the metal layer
is too thick, the local temperature is too low to form fully molten tiny catalyst balls,
or the metal is wetted well onto the substrate, and the catalyst remains at the root of
the nanowires. The ultra-small ZnO nanobelts reported recently are examples of this
case
21
(Figure 4.5(c)). The structure is catalyzed by a thin layer of tin that was precoated
on a silicon substrate. Using the VLS growth, random distributed, freestanding 1-D
nanostructures can also be obtained by evaporizing a mixture of the catalyst and source
materials. In this process, the formation of alloy droplets and the growth of 1-D
nanostructures occurred during vapor transport.
22,23
ZnO nanowires, nanobelts, and
nanorods have been grown using Sn as a catalyst, but Sn metal was not introduced
directly, but rather by reducing SnO
2
mixed in the source material.
FIGURE 4.5 One-dimensional nanostructure grown by the VLS process. (a) Aligned ZnO
nanorods growing on a sapphire substrate with gold catalyst at the tips; (b) ZnO nanowires
growing on a silicon substrate with gold catalyst at the tips; (c) ZnO nanobelts rooted at tin
catalysts.
50 nm
In VLS growth, the metal particles are usually believed to be in a liquid state
during the growth and its solid-state structure may have no influence on the nature
of the grown nanowires and nanobelts. Our recent study on the ZnO nanowire and
nanobelts using Sn as a catalyst indicates that the crystalline orientation of the Sn
particle can determine the growth direction and the side surfaces of the nanowires
and nanobelts.
23
Studies on the interface relationship between catalyst tin particles
and their guided ZnO 1-D nanostructures indicate that tin catalyst can guide not
only [0001] growth of nanowires, but also [ ] and [ ] growth of nanobelts.
The key is to minimize the interface mismatch energy. The results reveal that the
interfacial region of the tin particle with the ZnO nanowire/nanobelt could be
partially crystallized during the VLS growth, although the local growth temperature
is much higher than the melting point of tin. This result may have an important
impact on understanding the physical chemical process in the VLS growth.
4.5 THE NANOBELT FAMILY
Since the discovery of nanobelts in our laboratory in 2001, a wide range of interesting
nanostructures of functional oxides have been synthesized. Nanobelts represent a
unique group of nanostructures from semiconducting oxides to II-VI and III-V
compounds to structural ceramics. The family of nanobelts is now experiencing a
flourishing development and will be a new focus area in nanoscience and nano-
technology.
4.5.1 BINARY OXIDE NANOBELTS
Belt morphology was first discovered in semiconducting oxide materials. Table 4.1
summarizes the nanobelt structures of function oxides that have been discovered
experimentally thus far, including ZnO,
2
SnO
2
,
24
Ga
2
O
3
,
25
In
2
O
3
,
26
CdO,
2
PbO
2
,
27
TABLE 4.1
Crystallographic Geometry of Functional Oxide Nanobelts
Nanobelt
Crystal
Structure Growth Direction Top Surface Side Surface
ZnO Wurtzite [0001] or [ ] r( ) or r(2 0) r(01 0) or r(0001)
Ga
2
O
3
Monoclinic [001] or [010] r(100) or r(100) r(010) or r(10 )
t-SnO
2
Rutile [101] r(10 ) r(010)
o-SnO
2
Orthorhombic [010] r(100) r(001)
In
2
O
3
C-rare earth [001] r(100) r(010)
CdO NaCl [001] r(100) r(010)
PbO
2
Rutile [010] r(201) r(10 )
MoO
3
Orthorhombic [001] r(010) r(100)
MgO Face Center
Cubic
[100] r(001) r(010)
CuO Monoclinic [ 10] r(110) r(001)
Al
2
O
3
Hexagonal [001] r( 10) r(110)
0110 21 10
0110 21 10 1 1 1
1
1
1
1
1
MoO
3
,
28,29
MgO,
30,31
CuO,
32
and Al
2
O
3
.
33
Although these materials belong to dif-
ferent crystallographic families, they do have a common faceted structure, which
is the nanobelt structure. Each type of nanobelt is defined by its crystallographic
structure, the unique growth direction, and specific top/bottom surfaces and side
surfaces. Some of the materials can grow along different crystal directions, which
can be controlled by choosing specific experimental conditions.
4.5.2 COMPOUND SEMICONDUCTOR NANOBELTS
In addition to the oxide nanobelts that are studied intensively, more and more
function non-oxide nanobelts have been fabricated using various approaches. They
are mostly II-VI or III-V compounds that have shown interesting physical or chem-
ical properties in bulk or other nanomorphologies. The structure and structural
characteristics are summarized in Table 4.2. ZnS
34
and CdSe nanobelts,
35
II-VI
compounds with the wurtzite crystal structure, have been recently synthesized.
Analogous to ZnO nanobelts, these nanobelts exhibit polarized side surfaces that
induce anisotropic growth, forming a one-side structure with teeth, a so-called
nanosaw.
36
Some other important III-V semiconducting materials, such as GaN,
3638
GaP,
39
and AlN,
40
have also been reported to exhibit belt-like morphology. These
nanobelts are single crystal and have specified growth directions and side facets.
4.5.3 MULTIELEMENT NANOBELTS
Because of the high requirements of stoichiometry, the formation of multielement
nanobelts is restricted by the evaporation method, which is the traditional pathway
used to grow a nanobelt. However, through some chemical routes, a limited number
of three-component nanobelts can be fabricated in solution. For example, bismuth
oxide bromide
41
and BaCrO
4
42
nanobelts have been prepared through a hydrothermal
route and a reverse micelles directed synthesis, respectively.
Another type of nanobelt is achieved by doping. Using nanobelts as a template,
doping with another type of element results in a new type of nanostructure. Manga-
nese-doped ZnO nanobelts have been fabricated by ion implanting Mn
2+
into as-
synthesized ZnO nanobelts, which have very interesting applications in spintronics.
43
Nitrogen-doped GaP nanobelts have been achieved by growing GaP nanobelts in an
NH
3
atmosphere,
39
which shows interesting optical properties.
TABLE 4.2
Crystallographic Geometry of Functional Non-Oxide Nanobelts
Nanobelt Crystal Structure Growth Direction Top Surface Side Surface
ZnS Wurtzite [01 0] ( ) r(0001)
CdSe Wurtzite [01 0] ( ) r(0001)
GaN Hexagonal [10 0] r(01 0) r(0001)
GaP Zinc blende [111] r(100) r(0 1)
AlN Face Center Cubic [001] r(100) r(010)
1 21 10
1 21 10
1 1
1
4.6 ULTRA-NARROW ZnO NANOBELTS
In general, it is believed that nanostructures less than 10 nm in size will have novel
and unique physical and chemical properties due to quantum confinement. Nanowires
of a few nanometers in diameter have been synthesized for InP
44
and Si
45
by using
nanoparticles as catalysts or using single-wall carbon nanotubes as templates. How-
ever, due to the fact that solid-vapor phase synthesis typically occurs at high tem-
perature (between 700 and 1300qC), the synthesis of small-size (<10 nm) ZnO
nanobelts requires an innovative approach.
Using thin films of metallic tin as a catalyst, ultrasmall single-crystalline ZnO
nanobelts have recently been synthesized in large quantity by a carbon-thermal
evaporation process.
21
Instead of using presynthesized nanoparticles as catalysts, a
uniform thin film (~10 nm) of tin is coated on the silicon substrate. Since tin has a
very low melting point (232qC), it is molten and slowly oxidized
46
before the Zn/ZnO
vapor is formed in the reaction tube. During a VLS process, ultra-small ZnO nano-
belts grow out from the partially oxidized tin balls and cover the entire substrate
with a reasonable yield, as shown in Figure 4.6(a). Without using catalyst nanopar-
FIGURE 4.6 (a) Scanning electronic microscopy image of ultra-thin ZnO nanobelts grown
on tin coated on silicon substrate. (b) Size distribution of the as-synthesized ZnO nanobelts
measured from TEM images.
35
(b)
30
25
20
15
10
5
0
N
u
m
b
e
r
3 4 5 6 7 8 9 10
Nonobelt width (nm)
ticles to control the size of the nanobelts as in conventional VLS growth, this
technique provides a much simpler way to fabricate ultra-small ZnO nanobelts at
lower cost and in large quantity.
The nanobelts grow along the [0 0 01] direction, with ( ) top/bottom
surfaces and ( ) side surfaces. The average width of the nanobelts is only 5.5
nm with a narrow size distribution of r1.5 nm. As shown in Fig. 4.6(b), most
nanobelts are in the range of 47 nm, indicating very good size uniformity. All of
the nanobelts display a single crystalline structure with the (0001) fringes of spacing
5.1 . Unlike larger ZnO nanobelts that are dislocation free, the ultra-narrow
nanobelts have dislocations, which are revealed by high-resolution TEM. It is
assumed that the formation of dislocations is due to a large local strain introduced
by the bending of the nanobelts. This is demonstrated by the fact that a nanobelt
of 10 nm in width is straight (Fig. 4.7(a)), while a thin nanobelt of 5 nm in width
shows a curly shape (Fig. 4.7(b)). The strain in the thicker nanobelt is released
by creating stacking faults perpendicular to the growth direction of the nanobelts
(Fig. 4.7(c)). For the ~5-nm nanobelt, besides stacking faults, distortion in the orien-
tation of the (0001) atomic planes (Fig. 4.7(d)) and edge dislocations (Fig. 4.7(e))
FIGURE 4.7 High-resolution TEM images of (a) 10 nm wide and (b) 5 nm wide ZnO
nanobelts. (ce) Enlarged images from areas c, d, and e marked in (a) and (b) showing the
stacking fault, lattice distortion, and edge dislocation, respectively.
21 10
0110
are introduced to accommodate the local deformation. The interplanar distance can
be expanded to ~5.4 at the exterior arc and compressed to ~4.8 at the inner arc
to accommodate the local strain. This is possible for narrow nanobelts. The creation
of edge dislocations in oxide nanobelts is rare.
Photoluminescence (PL) measurements are performed to examine the size-
induced quantum effect in the ultra-thin ZnO nanobelts. Figure 4.8 compares the PL
spectra recorded from the ZnO nanobelts that have an average width of ~200 nm and
the PL from the ultra-thin ZnO nanobelts with an average width of 5.5 nm. The
387-nm peak corresponds to a 3.2-eV spontaneous emission of the normal ZnO
nanobelts. This emission lies at an energy considerably below that of the band gap
(3.37 eV) of ZnO, and is probably a result of exciton-exciton collision.
47
From this
model, the energy of the resulting photon can be given by
48
(n 2, 3, 4,)
where P
n
is the photon energy, E
ex
is the free-exciton emission energy, E
b
ex
is the
binding energy of the exciton, n is the quantum number of the envelope function,
and kT is the thermal energy. However, the ultra-thin ZnO nanobelts exhibit a near
band edge emission at 373 nm (3.32 eV), which was measured under identical
experimental conditions as for the larger nanobelts. It is believed that the increase
FIGURE 4.8 Photoluminescence spectra acquired from the ~200 nm-wide ZnO nanobelts
(blue) and the 6 nm-wide ZnO nanobelts (pink).
W = 200 nm
W = 6 nm
360 370 380 390 400
Wavelength (nm)
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
3
7
3
3
8
7
P E E
n
kT
n ex b
ex
=
1
1 2
3
2
in photon energy is due to an increase in E
ex
, which indicates quantum confinement
due to the reduced size of the nanobelts. The ultra-narrow ZnO nanobelts can be
used in solid-state quantum computing, nanometer-size memory devices, UV laser
diodes, etc. They could be a new type of quantum nanowire.
4.7 MESOPOROUS ZnO NANOWIRES
Porous materials have been used in a wide variety of applications in bioengineering,
catalysis, environmental engineering, and sensor systems owing to their high surface-
to-volume ratio. Normally most mesoporous structures are composed of amorphous
materials
49,50
and the porosity is achieved by solvent-based organic or inorganic
reactions.
51
Recently, a new wurtzite ZnO nanowire structure that is single crystal
but composed of mesoporous walls/volumes was fabricated in our laboratory.
52
The
porous nanowires have a hexagonal cross section with a variety of pore sizes from
10 to 50 nm. The typical length varies from 100 Pm to 1 mm and the diameter is
in the range of 50500 nm (Figure 4.9(a)).
The growth of this porous nanostructure is catalyzed by a thin film of metallic tin
as well, but the deposition temperature, which dictates the final morphology, is much
higher than that used for growing ultra-small nanobelts. Condensation nucleation is
FIGURE 4.9 Mesoporous single-crystal ZnO nanowires. (a) Porous nanowires; (b) partially
aligned porous nanorods; (c) junctions between porous nanowires; (d) rib-like porous nanorod
arrays.
also highly activated by this elevated local substrate temperature. Thus, a few more
complex structures are formed, such as partially aligned ZnO sprouts that formed
coral-like structures (Figure 4.9(b)), crossed structures between nanowires (Figure
4.9(c)), and rib-like short nanorod arrays (Fig. 4.9(d)). Each of these structures
exhibits a high porosity surface.
TEM studies are performed to determine the mechanism of the formation of
the porous structures. An electron diffraction pattern recorded from a porous nanow-
ire (Figure 4.10(a)) is shown in Figure 4.10(b), which presents two sets of patterns.
The brighter spots come from ZnO, which is the [ ] zone axis pattern, dis-
playing the nanowire axial direction as [0 0 01]. The weaker diffraction spots
correspond to a new Zn
2
SiO
4
phase that is sheathed on the surface of the nanowire
and has an epitaxial orientation relation with ZnO, which is clearly shown in a
high-resolution TEM picture in Fig. 4.10(c). The diffraction pattern from the
Zn
2
SiO
4
phase can be indexed to be the [100] zone axis. The orientation relationship
between ZnO and Zn
2
SiO
4
is (0001)
ZnO
|| , [ ]
ZnO
|| ,
where structurally Zn
2
SiO
4
has an orthorhombic crystal structure with lattice constants
of a 0.9085 nm, b 1.0625 nm, and c 0.5962 nm.
Therefore it is believed that the formation of this porous structure is due to the
appearance of the Zn
2
SiO
4
out-layer. In a high temperature region, ZnO nanowires
can redecompose into Zn vapor and O
2
, which could happen at the local substrate
temperature of ~600qC. In the meantime, Si-O vapor sublimated from the silicon
FIGURE 4.10 (a) Low-magnification TEM image of a porous ZnO nanowire and (b) corre-
sponding diffraction pattern; (c) high-magnification TEM image taken on the edge of the
porous ZnO nanowire.
21 10
( ) 001
Zn SiO
2 4
21 10 [ ] 100
Zn SiO
2 4
substrate can be quickly deposited on the nanowire surface and diffuse into the ZnO
lattice, resulting in the formation of a new phase Zn
2
SiO
4
. The electronegativity of
Si is 1.9 and the electronegativity of Zn is 1.65, which are quite close. The atomic
sizes of Si and Zn are 0.117 and 0.133 nm, respectively, which are comparable. Zinc
oxide and Si-O vapor can quickly combine to form a new phase:
The newly formed Zn
2
SiO
4
layer tends to have an epitaxial relationship with the
ZnO in order to reduce the interface lattice mismatch. On the other hand, the lattice
mismatch between (0001)
ZnO
and is 14%, and between ( )
ZnO
and
4 is 9%, and Zn
2
SiO
4
tends to form textured islands on the surface of
ZnO, but they cannot form a continuous single-crystal film due to large mismatch,
resulting in the growth of epitaxial Zn
2
SiO
4
islands on the ZnO surface. The newly
formed Zn
2
SiO
4
may be more stable than ZnO at the local growth temperature.
Therefore, there are open areas on the ZnO surface that are not covered by the
Zn
2
SiO
4
network; an evaporation of Zn-O from the open areas may form the pores
in the volume of the nanowire. The sublimation of ZnO and growth of Zn
2
SiO
4
proceed simultaneously and finally end with a high-porosity ZnO nanostructure.
This porous structure not only exhibits a high surface-to-volume ratio, but also
provides numerous holes that are big enough to store other functional nanoparticles.
Meanwhile the mesoporous ZnO nanowires still retain the crystallized structure, and
their unique electronic and optical properties are preserved. The high surface-to-
volume ratio guarantees a large activation area and the conductivity enables the
signal to be easily transported. Therefore, the mesoporous ZnO nanowires are good
candidates for sensor systems. The mesoporous ZnO nanowires can also be used in
electrochemical reactions as electrodes that hold catalysts and provide a large reac-
tion area.
4.8 PATTERNED GROWTH OF ALIGNED
ZnO NANOWIRES
Aligned growth of ZnO nanorods/nanowires has been successfully achieved on a
solid substrate via a VLS process with the use of gold
53,54
and tin
14
as catalysts. The
catalyst initiates and guides the growth, and the epitaxial orientation relationship
between the nanorods and the substrate leads to the aligned growth. Other techniques
that do not use any catalyst, such as metalorganic vapor-phase epitaxial growth,
55
template-assisted growth,
4
and electrical field alignment,
56
have also been employed
for the growth of vertically aligned ZnO nanorods. Based on a gold catalyst template
produced by a self-assembled monolayer of submicron spheres and guided VLS
growth on a single crystal alumina (sapphire) substrate, aligned ZnO nanorods have
been hexagonally patterned in a large area.
57
The synthesis process involves three main steps. First a 2-D, large-area, highly
ordered monolayer of submicron polystyrene (PS) spheres was self assembled on
2
2
600
2 4
ZnO SiO Zn SiO
C
+
~
( ) 001
Zn SiO
2 4
0110
( ) 010
Zn SiO
2 4
a single crystal Al
2
O
3
substrate (Figure 4.11(a)). This can be achieved in many
different ways.
5859
Second, the self-assembled arrays of PS spheres were then used
to pattern the gold catalyst. For this process, gold particles were either sputtered or
thermally evaporated onto the self-assembled monolayer structure; as a result, a
honeycomb-like hexagonal gold pattern and a highly ordered hexagonal array of
gold spots pattern were obtained, respectively (Figures 4.11(b), (c)). Finally, using
a carbon-thermal evaporation method, aligned ZnO nanorods are obtained through
a VLS process. Following the patterned gold catalyst, the aligned ZnO nanorods
exhibit a hexagonal arrangement on the sapphire substrate (Figure 4.11(d)).
Clearly the ZnO nanorods are oriented perpendicular to the sapphire substrate,
with a gold particle at the tip. All of the ZnO nanorods have about the same height,
of about 1.5 Pm, and their diameters range between 50 and 150 nm. By changing
the growth time, the height of the ZnO nanorods could be varied from a few hundred
nanometers to a few micrometers.
The patterned and aligned 1-D ZnO nanostructures hold huge promise for applica-
tions as sensor arrays, piezoelectric antenna arrays, optoelectronic devices, and inter-
connects. Angular-dependent photoluminescence spectra showed that the luminescence
was emitted mainly along the axis of the ZnO nanorods. As shown in Figure 4.12, by
increasing the detection angle, the luminescence intensity dropped dramatically. More-
over the luminescence peak shifted very slightly from 384 to 383 nm when the
FIGURE 4.11 (a) Self-assembled monolayer of polystyrene spheres; (b) gold catalyst pattern
achieved by sputtering; (c) gold catalyst pattern achieved by thermal evaporation; (d) aligned
ZnO nanorods grown on substrate.
detection angle increased from 0q to 45q. This may be caused by the polarization
of the emitted light from the aligned nanorods.
60
Thus a very tiny change of the
optical or electrical property can be magnified by the aligned structure instead of
canceling out by a random distribution so that the sensitivity could be highly
increased and could be applied in a macrosensing system.
FIGURE 4.12 (a) Photoluminescence (PL) spectra acquired from an aligned ZnO nanorod
array as a function of the angle between the detector and a direction normal to the substrate;
(b) the experimental setup.
384 nm
383 nm
0
15
30
45
360 370 380 390 400 410 420 430
Wavelength (nm)
I
n
t
e
n
s
i
t
y

(
a
.
u
.
)
(a)
(b)
Detecting angle
Laser
Substrate
Detector
45
4.9 SELECTED APPLICATIONS OF NANOBELTS
4.9.1 NANOCANTILEVERS
Cantilever-based scanning probe microscopy (SPM) is a powerful imaging, manip-
ulating, and measuring technique in the nanoscale. An ideal SPM cantilever to detect
any forces between the tip and sample would be a robust 1-D structure that is small
and rigid. The SPM cantilevers are conventionally made from Si, SiC, and Si
3
N
4
,
with a typical size of a125, a35, and a4 Pm in length, width, and thickness,
respectively. Recently upon the discovery of semiconducting oxide nanobelts, it was
found that nanobelts could be ideal candidates for cantilever applications.
Structurally, most of the nanobelts are single crystals and dislocation free, which
provides them excellent mechanical properties. A ZnO nanobelt of 56 nm in width can
be bent for more than 60q without breaking, thus demonstrating extremely high mechan-
ical flexibility. The elastic modulus of ZnO nanobelts has been measured by an in situ
TEM technique based on electromechanical resonance. Geometrically, depending on
the synthesis conditions, the dimensions of nanobelts vary from a6 to 300 nm, which
is about 35 to 1800 times smaller than the conventional cantilevers. The greatly reduced
cantilever sizes will significantly increase the physical, chemical, and biological sensi-
tivity of SPM cantilevers.
There are feasible methods of manipulating the nanobelts so as to realize the
cantilever applications.
61
Due to the capillary force between the atomic force micro-
scope (AFM) probe and the nanobelt, an individual ZnO nanobelt can be picked up
selectively by an AFM probe and cut into specific lengths if the technician keeps
bending the nanobelt using the AFM tip. Combining these techniques, individual
ZnO nanobelts with different lengths have been aligned horizontally onto a silicon
wafer (Figure 4.13(a)). This exemplifies our ability to tune the resonance frequency
of each and to modify cantilevers for different applications such as contact, noncon-
tact, and tapping mode AFM.
The nanocantilever can be applied to detect molecules (Figure 4.13(b)). When
a small number of target molecules are absorbed onto the cantilever, the total mass
of the cantilever will be changed. This small change can be magnified by shifting
the resonance frequency, which can be applied to detect molecules at the sensitivity
of a single molecule depending on the quality and size of the cantilever.
4.9.2 FIELD-EFFECT TRANSISTORS
By contacting the nanostructure to metal electrodes, a field-effect transistor (FET)
can be produced, which allows the exploration of the electrical properties of the
nanostructures. Based on single SnO
2
and ZnO nanobelts, FETs have been fabricated
using two different means.
62
A SnO
2
FET is formed by depositing a nanobelt onto
a SiO
2
/Si(P
+
) substrate, which is followed by making Ti electrodes using e-beam
lithography. A ZnO FET is made by directly applying ethanol-dispersed nanobelts
on predefined gold electrode arrays (Figure 4.14(a)). In both cases, a gold layer
evaporated on the Si (P
) side of the substrate is used as the back gate electrode,

the control of which reveals the current-voltage (I-V) characteristics of the FETs.
Figure 4.14(b) demonstrates the I-V properties of the FET based on a ZnO nanobelt,
showing a gate threshold voltage of 15 V, a switching ratio of nearly 100, and a
peak conductivity of 1.25 10
3
( cm)
1
. Furthermore a ZnO nanobeltbased FET
can be controlled by UV illumination as well. When a UV light with a wavelength
shorter than the band gap of ZnO is introduced, an immediate increase in the source-
drain current can be observed, indicating a switch of status from OFF to ON.
According to the measurements, the nanobelt FETs have switching ratios as large
as six orders of magnitude, conductivity as high as 15 ( cm)
1
, and electron
mobility as large as 125 cm
2
/V s.
FIGURE 4.13 Nanobelt as nanosensor. (a) ZnO nanobelts of different lengths aligned on
silicon substrate as nanocantilevers. The insets are the enlarged SEM image of the nanobelts.
(b) A schematic setup of a cantilever-based molecular detector (left) and its working mech-
anism (right).
(b)
Target
molecule
Probe molecule
Cantilever
beam
A
m
p
l
i
t
u
d
e
Frequency
Chip
(a)
30 m
FIGURE 4.14 Field effect transistor using a nanobelt. (a) The experimental setup for meas-
uring the electronic properties of ZnO nanobelts; (b) source-drain current vs. gate bias for a
ZnO field-emission transistor.
Gold electrodes (50 nm thick)
SiO
2
(~120 nm thick)
10 m
ZnO nanoribbon
(w = 285 nm, h = 110 nm)
(a)
ON
OFF
Vds = 6.5 V
20 10 0 10 20
Backgate potential (V)
(b)
10
7
10
8
10
9
10
10
S
o
u
r
c
e
-
d
r
a
i
n

c
u
r
r
e
n
t

(
A
)
8
6
4
2
8
6
4
2
8
6
4
2
4.9.3 NANOSENSORS
Semiconductor thin films are the most promising devices among solid-state chemical
sensors due to their low cost, low power consumption, and high compatibility.
However, in a thin-film device, which is normally polycrystalline, only a small
fraction of the species absorbed near the grain boundaries is active in modifying the
conductivity. Semiconducting metal oxide nanobelts are defect-free single crystals
with a very high surface-to-volume ratio due to their small sizes. These features
provide a possible high sensitivity because the faces exposed to the gaseous envi-
ronment are always the same and the small size is likely to produce a complete
depletion of carriers into the nanobelt, which will change the electrical properties.
SnO
2
nanobelts have been successfully fabricated as gas sensors
63
by placing
the nanobelts on platinum electrodes, where the contact is confirmed to be ohmic
by I-V measurements. Prompt responses of the current change are observed when
this gas sensor is exposed to a gas flow with a small amount of certain species such
as CO, NO
2
, and ethanol. As shown in Figure 4.15, the current jumps up immediately
FIGURE 4.15 Nanosensors using nanobelts. Response of SnO
2
sensors to CO, ethanol, and
NO
2
gases at 200qC.
10
5
10
6
10
7
0 50 100 150 200
Time (min)
1000
100
10
1
0.1
T 200C
C
u
r
r
e
n
t

(
A
)
C
o
n
c
e
n
t
r
a
t
i
o
n

(
p
p
m
)
CO (ppm)
Ethanol (ppm)
NO
2
(ppm)
once a few hundreds ppm CO or ethanol is introduced into the system, while only
0.5 ppm NO decreases the conductivity of the SnO nanobelts, obviously. When the
gas species are depleted, the current recovers to the original level indicating a good
repeatability of the SnO
2
nanobelts sensitivity. The results demonstrate the feasi-
bility for fabricating highly sensitive nanosized gas sensors using the integrity of a
single nanobelt.
4.10 SUMMARY
This chapter provides an overview of the synthesis, structure, and potential appli-
cations of functional oxide nanobelts. Due to their structurally controlled morphol-
ogy, nanobelts represent a new group of 1-D nanostructures. The most popular
technique for synthesizing nanobelts is by vapor deposition with or without metal
catalysts through the vapor-solid or vapor-liquid-solid (VLS) process, respectively.
By using Sn and Au as catalysts, some new 1-D ZnO nanostructures have been
synthesized. Ultra-narrow ZnO nanobelts with an average width of 5.5 nm exhibit
a clear blue shift in photoluminescence due to size effect. Mesoporous ZnO nanow-
ires with a pore size range from 10 to 50 nm could be good candidates for chemical
and environmental sensors. By patterning gold catalyst through self-assembled sub-
micron sphere arrays, aligned ZnO nanorods with a hexagonal arrangement are
fabricated that exhibit great potential applications for sensor systems and optoelec-
tronic devices. Owing to their well-controlled structures and unique chemical, phys-
ical, and electrical properties, semiconducting nanobelts could be applied in a broad
area, including nanocantilevers, sensors, transducers, FETs, resonators, and more.
We expect research in functional oxide to reach a new horizon in the next few years.
PROBLEMS
1. How can a nanobelt structure be defined? What are the differences
between nanowires, nanorods, and nanobelts?
2. List at least five methods to synthesize 1-D nanostructures. What are the
basic procedures for the thermal evaporation technique? What are the
advantages and disadvantages of the thermal evaporation technique? How
can the disadvantages be overcome?
3. Describe how 1-D nanostructures form through a vapor-liquid-solid (VLS)
process. How does a metal catalyst affect the final morphologies of 1-D
nanostructures in a VLS process?
4. Describe the mechanism of the formation of the porous ZnO 1-D nano-
structure.
5. What are the three main steps for fabricating the patterned and aligned
ZnO nanorods?
6. List the possible applications of nanobelts. What are the advantages of
using a nanobelt as an SPM cantilever?
ACKNOWLEDGMENTS
The results in this chapter were partially contributed by our group members and
collaborators. We are grateful to Yong Ding, William Hughes, M.S. Arnold, Ph.
Avouris, E. Comini, G. Faglia, and G. Sberveglieri. Research was supported by NSF,
NASA, and DARPA.
REFERENCES
1. Wang, Z.L., Nanowires and Nanobelts, Vol. I: Metal and Semiconductor Nanowires,
and Vol. II: Nanowire and Nanobelt of Functional Oxide, Kluwer Academic Publish-
ers Dordrecht, the Netherlands, 2003.
2. Pan, Z.W., Dai, Z.R., and Wang, Z.L., Nanobelts of semiconducting oxides, Science,
291, 19471949, 2001.
3. Morales, A.M. and Lieber, C.M., A laser ablation method for the synthesis of crys-
talline semiconductor nanowires, Science, 279, 208211, 1988.
4. Liu, C., Zapien, J.A., Yao, Y., Meng, X., Lee, C.S., Fan, S., Lifshitz, Y., and Lee,
S.T., High-density, ordered ultraviolet light-emitting ZnO nanowire arrays, Adv.
Mater., 15, 838841, 2003.
5. Wang, X.D., Gao, P.X., Li, J., Summers, C.J., and Wang, Z.L., Rectangular porous
ZnO-ZnS nanocables and ZnS nanotubes, Adv. Mater., 14, 17321735, 2002.
6. Choi, Y.C., Kim, W.S., Park, Y.S., Lee, S.M., Bae, D.J., Lee, Y.H., Park, G.S., Choi,
W.B., Lee, N.S., and Kim, J.M., Catalytic growth of E-Ga
2
O
3
nanowires by arc
discharge, Adv. Mater., 12, 746750, 2000.
7. Ma, C., Berta, Y., and Wang, Z.L., Patterned aluminum nanowires produced by
electron beam at the surface of AlF
3
single crystal, Solid State Commun., 129,
681685, 2004.
8. Krumeich, F., Muhr, H.J., Niederberger, M., Bieri, F., Schnyder, B., and Nesper, R.,
Morphology and topochemical reactions of novel vanadium oxide nanotubes, J. Am.
Chem. Soc., 121, 83248331, 1999.
9. Dai, Z.R., Pan, Z.W., and Wang, Z.L., Novel nanostructures of functional oxides
synthesized by thermal evaporation, Adv. Funct. Mater., 13, 924, 2003.
10. Pan, Z.W., Xie, S.S., Chang, B.H., Wang, C.Y., Lu, L., Liu, W., Zhou, W.Y., Li, W.Z.,
and Qian, L.X., Very long carbon nanotubes, Nature, 394, 631632, 1998.
11. Cao, H., Xu, J.Y., Zhang, D.Z., Chang, S.H., Ho, S.T., Seelig, E.W., Liu, X., and
Chang, R.P.H., Spatial confinement of laser light in active random media, Phys. Rev.
Lett., 84, 5584, 2000.
12. Huang, M.H., Wu, Y., Feick, H., Tran, N., Weber, E., and Yang, P., Catalytic growth
of zinc oxide nanowires by vapor transport, Adv. Mater., 13, 113116, 2001.
13. Gudiksen, M.S., Wang, J., and Lieber, C.M., Synthetic control of the diameter and
length of single crystal semiconductor nanowires, J. Phys. Chem. B., 105, 40624064,
2001.
14. Gao, P.X., Yong, D., and Wang, Z.L., Crystallographic-orientation aligned ZnO nan-
orods grown by tin catalyst, Nano Letters, 3, 13151320, 2003.
15. Gao, P.X. and Wang, Z.L., Self-assembled nanowire-nanoribbon junction arrays of
ZnO, J. Phys. Chem. B., 106, 1265312658, 2002.
16. Li, S.Y., Lee, C.Y., and Tseng, T.Y., Copper-catalyzed ZnO nanowires on silicon
(1 0 0) grown by vaporliquidsolid process, J. Crystl. Growth, 247, 357362, 2003.
17. Duan, X. and Lieber, C.M., Laser-assisted catalytic growth of single crystal GaN
nanowires, J. Am. Chem. Soc., 122, 188189, 2000.
18. Dai, Z.R., Pan, Z.W., and Wang, Z.L., Gallium oxide nanoribbons and nanosheets,
J. Phys. Chem. B, 106, 902904, 2002.
19. Wagner, R.S. and Ellis, W.C., Vapor-liquid-solid mechanism of single crystal growth,
Appl. Phys. Lett., 4, 8990, 1964.
20. Huang, M.H., Mao, S., Feick, H., Yan, H., Wu, Y., Kind, H., Weber, E., Russo, R.,
and Yang, P. Room-temperature ultraviolet nanowire nanolasers, Science, 292,
18971899, 2001.
21. Wang, X.D., Ding, Y., Summers, C.J., and Wang, Z.L., Large-scale synthesis of
six-nanometer-wide ZnO nanobelts, J. Phys. Chem. B, 108, 87738777, 2004.
22. Duan, X. and Lieber, C.M., General synthesis of compound semiconductor nanowires,
Adv. Mater., 12, 298302, 2000.
23. Ding, Y., Gao, P.X., and Wang, Z.L., Catalyst-nanostructure interfacial lattice mis-
match in determining the shape of VLS grown nanowires and nanobelts: A case of
Sn/ZnO, J. Am. Chem. Soc., 126, 20662072, 2004.
24. Dai, Z.R., Gole, J.L., Stout, J.D., and Wang, Z.L., Tin oxide nanowires, nanoribbons,
and nanotubes, J. Phys. Chem. B., 106, 12741279, 2001.
25. Dai, Z.R., Pan, Z.W., and Wang, Z.L., Gallium oxide nanoribbons and nanosheets,
J. Phys. Chem. B, 106, 902904, 2002.
26. Kong, X.Y. and Wang, Z.L., Structures of indium oxide nanobelts, Solid State Com-
mun., 128, 14, 2003.
27. Pan, Z.W., Dai, Z.R., and Wang, Z.L., Lead oxide nanobelts and phase transformation
induced by electron beam irradiation, Appl. Phys. Lett., 80, 309311, 2002.
28. Li, Y.B., Bando, Y., Golberg, D., and Kurashima, K., Field emission from MoO
3
nanobelts, Appl. Phys. Lett., 81, 50485050, 2002.
29. Li, X.L., Liu, J.F., and Li, Y.D., Low-temperature synthesis of large-scale single-crystal
molybdenum trioxide (MoO
3
) nanobelts, Appl. Phys. Lett., 81, 48324834, 2002.
30. Li, Y.B., Bando, Y., and Sato, T., Preparation of network-like MgO nanobelts on Si
substrate, Chem. Phys. Lett., 359, 141145, 2002.
31. Liu, J., Cai, J., Sun, Y.C., Gao, Q.M., Suib, S.L., and Aindow, M., Magnesium
manganese oxide nanoribbons: Synthesis, characterization, and catalytic application,
J. Phys. Chem. B, 106, 97619768, 2002.
32. Wen, S.G., Zhang, W.X., and Yang, S.H., Synthesis of Cu(OH)2 and CuO nanoribbon
arrays on a copper surface, Langmuir, 19, 58985903, 2003.
33. Fang, X.S., Ye, C.H., Peng, X.S., Wang, Y.H., Wu, Y.C., and Zhang, L.D., Temperature-
controlled growth of D-Al
2
O
3
nanobelts and nanosheets, J. Mater. Chem., 13,
30403043, 2003.
34. Ma, C., Moore, D., Li, J., and Wang, Z.L., Nanobelts, nanocombs and nano-windmills
of wurtzite ZnS, Adv. Mater., 15, 228231, 2003.
35. Ma, C., Ding, Y., Moore, D., Wang, X.D., and Wang, Z.L., Single-crystal CdSe
nanosaws, J. Am. Chem. Soc., 126, 708709, 2004.
36. Wang, Z.L., Kong, X.Y., and Zuo, J.M., Induced growth of asymmetric nanocantilever
arrays on polar surfaces, Phys. Rev. Lett., 91, 185502, 2003.
37. Xiang, X., Cao, C.B., and Zhu, H., Catalytic synthesis of single-crystalline gallium
nitride nanobelts, Solid State Commun., 126, 315318, 2003.
38. Bae, S.Y., Seo, H.W., Park, J., Yang, H., and Song, S.A., Synthesis and structure of
gallium nitride nanobelts, Chem. Phys. Lett., 365, 525529, 2002.
39. Seo, H.W., Bae, S.Y., Park, J., Yang, H., Kang, M., Kim, S., Park, J.C., and Lee, S.Y.,
Nitrogen-doped gallium phosphide nanobelts, Appl. Phys. Lett., 82, 37523754, 2003.
40. Wu, Q., Hu, Z., Wang, X.Z., Chen, Y., and Lu, Y.N., Synthesis and optical charac-
terization of aluminum nitride nanobelts, J. Phys. Chem. B, 107, 97269729, 2003.
41. Wang, J. and Li, Y., Synthesis of single-crystalline nanobelts of ternary bismuth oxide
bromide with different compositions, Chem. Comm., 18, 23202321, 2003.
42. Shi, H.T., Qi, L.M., Ma, J.M., Cheng, H.M., and Zhu, B.Y., Synthesis of hierarchical
superstructures consisting of BaCrO
4
nanobelts in catanionic reverse micelles, Adv.
Mater., 15, 16471651, 2003.
43. Ronning, C., Gao, P.X., Ding, Y., Wang, Z.L., and Schwen, D., Manganese doped
ZnO nanobelts for spintronics, Appl. Phys. Lett., 84, 783785, 2004.
44. Gudiksen, M.S., Wang, J., and Lieber, C.M., Synthetic control of the diameter and
length of single crystal semiconductor nanowires, J. Phys. Chem. B, 105, 40624064,
2001.
45. Ma, D.D.D., Lee, C.S., Au, F.C.K., Tong, S.Y., and Lee, S.T., Small-diameter silicon
nanowire surfaces, Science, 299, 18741877, 2003.
46. Huh, M., Kim, S., Ahn, J., Park, J., and Kim, B., Oxidation of nanophase tin particles,
NanoStructured Mater., 11, 211220, 1999.
47. Bagnall, D.M., Chen, Y.F., Zhu, Z., Yao, T., Koyama, S., Shen, M.Y., and Goto, T.,
Optically pumped lasing of ZnO at room temperature, Appl. Phys. Lett., 70,
22302232, 1997.
48. Klingshirn, C., Luminescence of ZnO under high one-quantum and 2-quantum exci-
tation, Phys. Status Solidi B, 71, 547556, 1975.
49. Kramer, E., Forster, S., Goltner, C., and Antonietti, M., Synthesis of nanoporous silica
with new pore morphologies by templating the assemblies of ionic block copolymers,
Langmuir, 14, 20272031, 1998.
50. Hornebecq, V., Mastai, Y., Antonietti, M., and Polarz, S., Redox behavior of nano-
structured molybdenum oxide-mesoporous silica hybrid materials, Chem. Mater., 15,
35863593, 2003.
51. Cooper, A.I., Porous materials and supercritical fluids, Adv. Mater., 15, 10491059,
2003.
52. Wang, X.D., Summers, C.J., and Wang, Z.L., Mesophorous single-crystal ZnO nano-
wires epitaxially sheathed with Zn, SiO
4
, Adv. Mater., 16, 12151218, 2004.
53. Ng, H.T., Chen, B., Li, J., Han, J., Meyyappan, M., Wu, J., Li, X., and Haller, E.E.,
Optical properties of single-crystalline ZnO nanowires on m-sapphire, Appl. Phys.
Lett., 82, 20232025, 2003.
54. Zhao, Q.X., Willander, M., Morjan, R.R., Hu, Q.H., and Campbell, E.E.B., Optical
recombination of ZnO nanowires grown on sapphire and Si substrates, Appl. Phys.
Lett., 83, 165167, 2003.
55. Park, W.I., Kim, D.H., Jung, S.W., and Yi, G.C., Metalorganic vapor-phase epitaxial
growth of vertically well-aligned ZnO nanorods, Appl. Phys. Lett., 80, 42324234, 2002.
56. Harnack, O., Pacholski, C., Weller, H., Yasuda, A., and Wessels, J.M., Rectifying
behavior of electrically aligned ZnO nanorods, Nano Lett., 3, 10971101, 2003.
57. Wang, X.D., Summers, C.J., and Wang, Z.L., Large-scale hexagonal-patterned growth
of aligned ZnO nanorods for nano-optoelectronics and nanosensor arrays, Nano Lett.,
4, 423426, 2004.
58. Kempa, K., Kimball, B., Rybczynski, J., Huang, Z.P., Wu, P.F., Steeves, D., Sennett,
M., Giersig, M., Rao, D.V.G.L.N., Carnahan, D.L., Wang, D.Z., Lao, J.Y., Li, W.Z.,
and Ren, Z.F., Photonic crystals based on periodic arrays of aligned carbon nanotubes,
Nano Lett., 3, 1318, 2003.
59. Garno, J.C., Amro, N.A., Mesthrige, W., and Liu, G.Y., Production of periodic arrays
of protein nanostructures using particle lithography, Langmuir, 18, 81868192, 2002.
60. Johnson, J.C., Yan, H., Yang, P., Saykally, R.J., Optical cavity effects in ZnO nanowire
lasers and waveguides, J. Phys. Chem. B, 107, 88168828, 2003.
61. Hughes, W. and Wang, Z.L., Nanobelt as nanocantilever, Appl. Phys. Lett., 82,
28862888, 2003.
62. Arnold, M.S., Avouris, P., Pan, Z.W., and Wang, Z.L., Field-effect transistors based
on single semiconducting oxides nanobelts, J. Phys. Chem. B, 107, 659663, 2003.
63. Comini, E., Faglia, G., Sberveglieri, G., Pan, Z.W., and Wang, Z.L., Stable and high-
sensitive gas sensors based on semiconducting oxide nanobelts, Appl. Phys. Lett., 81,
18691871, 2002.
5
Advances in Chemical
Vapor Deposition
of Carbon Nanotubes
Vesselin N. Shanov, Atul Miskin, Sachin Jain,
Peng He, and Mark J. Schulz
CONTENTS
5.1 The CVD Technique for Growth of CNT
5.2 CVD Growth System
5.3 Catalyst and Substrate Preparation
5.4 Growth of CNT
5.5 Purification of As-Grown CNT
5.6 Characterization of CNT
5.6.1 Scanning Electron Microscopy
5.6.2 Transmission Electron Microscopy
5.6.3 Raman Spectroscopy
5.7 Advanced Topics and Future Directions for CVD of CNT
5.7.1 Synthesis of Long Carbon Nanotubes
5.7.2 Design and Fabrication of a Segment for CNT Actuator
5.7.3 Oriented Growth of a Simple CNT Network
5.7.4 Functionalization of CNT
5.8 Conclusions
Problems
Acknowledgments
References
5.1 THE CVD TECHNIQUE FOR GROWTH OF CNT
In recent years there has been an enormous number of publications on carbon nanotube
(CNT) synthesis. Despite this information flow many challenges still need to be
addressed. One such challenge is the production of large-scale and low-cost single-
wall (SW) and multiwall (MW) carbon nanotubes. Another field of interest is the
pursuit of controlled CNT growth in terms of selective deposition, orientation, and
preselected metallic or semiconducting properties. Our understanding of the CNT
growth mechanism has been rapidly evolving, but more consideration is still required
to explain the variety of the observed growth features and experimental results. A
significant amount of information covering these aspects of CNT research, along with
other related topics, can be found in References 15.
Currently there are three principal techniques to produce high-quality SWCNT:
laser ablation,
6
electric arc discharge,
7,8
and chemical vapor deposition (CVD).
9,10
Laser ablation and arc discharge are modified physical vapor deposition (PVD) tech-
niques that involve the condensation of hot gaseous carbon atoms generated from the
evaporation of solid carbon. However, the equipment requirements and the large
amount of energy consumed by these methods make them mostly suitable for
laboratory research.
9,10
Both the laser ablation and arc-discharge techniques are
limited in the volume of sample they can produce in comparison to the size of the
carbon source. In addition, more impurities accompany the nanotubes in the form
of amorphous carbon and catalyst particles because of the high-temperature nature
of the heat source. Since the growth is difficult to control, the final product consists
mostly of MWCNT with poor alignment.
Chemical vapor deposition has become the most important commercial
approach for manufacturing of carbon nanotubes. CVD is known as irreversible
deposition of a solid from a gas or a mixture of gases through a heterogeneous
chemical reaction. This reaction takes place at the interface of gassolid substrate,
and depending on the deposition conditions, the growth process can be controlled
either by diffusion or by surface kinetics. CVD is the preferred technique for
fabrication of thin layers of metals, insulators, and semiconductors on different
substrates.
1113
This method can easily be scaled up to industrial production. CVD
is a continuous process that is currently the best-known technique for high-yield
and low-impurity production of CNT at moderate temperatures. It offers better
growth control because of the equilibrium nature of the chemical reactions involved.
In addition, CVD has the ability to control the size, shape, and alignment of the
nanotubes.
10
CVD of CNT uses carbon precursor gases such as benzene, methane, acetylene,
carbon monoxide, or ethanol. The process usually involves high-temperature decom-
position of hydrocarbons in hydrogen over the catalyst, which is predeposited on
the solid substrate. Figure 5.1 illustrates the CVD process by showing a typical
reactor and thermal decomposition of methane precursor in a hydrogen atmosphere.
The decomposition reaction can be controlled by the temperature, pressure, or
concentration of the reactants involved. This reaction generates carbon atoms, which
reach the surface of the substrate, where they diffuse into tiny islands of melted
catalyst that dissolve the carbon atoms, forming a metal-carbon solution. This high-
temperature solution may become supersaturated. When it reaches this point, carbon
precipitates, thus providing the building bricks for growth of CNT structures.
Transition metals such as Fe, Co, and Ni are frequently used as catalysts. The metal
catalyst can stay on top of CNT, forming a cap that encapsulates the CNT.
14
In some
cases, it remains at the bottom of the CNT, which promotes a bamboo-shaped
morphology.
15
The described growth mechanism is very simplistic; it postulates
that carbon atoms from the gas phase are transported through the melted carbon-
catalyst solution to the growth surface of the CNT. This process is based on the
vaporliquidsolid (VLS) mechanism. The VLS mechanism is used to explain crys-
tal growth of whiskers from a gas phase through molten (liquid) media located at
the gassolid interface.
16
The selection and preparation of the substrate and catalyst
are important aspects affecting the nanotube yield and the ability to purify the product
of the synthesis.
The CVD synthesis of CNT can be divided into two major groups: bulk syn-
thesis and surface synthesis on flat substrates.
10
The first group is based on ceramic
particles such as Al
2
O
3
or MgO, in some cases mixed with SiO
2
to form hybrid
materials.
1721
Catalysts of pure metals or alloys are deposited on the ceramics,
which provide support and structural stability at elevated temperatures. High-sur-
face-area catalysts with a porous structure of the supported ceramics are desirable
for high yield of CNT grown by CVD. Alumina-supported Fe-Mo is a frequently
used substrate/catalyst combination for synthesis of SWCNT.
9
Surface synthesis of SWNT requires flat substrates such as polished or porous Si
wafers, patterned with metal catalysts by various lithographic techniques.
2227
One
advantage of the surface synthesis is that the final CNT product is much cleaner than
the one grown by bulk synthesis. In addition, this technique provides better control of
the diameter and the alignment of the carbon nanotubes, which makes it attractive for
device fabrication. The size and the pattern of the metalcatalyst islands on the Si
substrate predetermine their orientation, geometric dimensions, and selective growth.
There are two different pathways to synthesize CNT by chemical vapor deposition.
The first one, as previously described, involves a solid catalyst supported by a substrate.
The second pathway requires the catalyst to be present in the vapor phase, where the
growth of CNT takes place. This is known as the floating catalyst approach, which
provides nanosize particles of the metal catalyst in the growth zone.
2831
Metallocenes
are used as a source of the catalyst precursor for the floating catalyst method. A typical
FIGURE 5.1 Schematic of a CVD reactor and the chemical reaction involved during the
growth of CNT.
Precursors
CH
4
H
2
Quartz
boat
Quartz
tube
Gas out
Gas inlet
Substrate:
Si wafer or
ceramic powder
Catalyst:
Fe, or Co, or Ni
CH
4
(g)
T
C(s) + 2H
2
(g)
H
2
floating catalyst compound that has been studied extensively is ferrocene.
2931
Less
popular are metallocenes such as cobaltocene.
31
Usually a solution of ferrocene in
benzene is introduced into the CVD reactor, either as an aerosol or as a vapor along
with a carrier gas. The reaction takes place at about 1100qC, depending on the stability
of the hydrocarbon. The role of ferrocene is to provide the iron catalyst, while benzene
delivers the carbon feedstock required for the CNT growth. Addition of sulfur in the
form of thiophene has been known to increase the yield of carbon nanotubes and is
used along with the catalytic metallocenes.
31
5.2 CVD GROWTH SYSTEM
Conventional CVD reactors are frequently used for growing CNT from carbon
precursors. An EasyTube
TM
Nanofurnace, shown in Fig. 5.2(a), is used at the
University of Cincinnati (UC) for the synthesis of carbon nanotubes.
32
The
furnace utilizes a thermally driven CVD process and consists of two main units:
the control unit and the process unit. The process unit shown in Fig. 5.2(b) is
composed of a furnace and a quartz tube reactor with a loader. The control unit
contains the workstation of the operator; its display is shown in Fig. 5.2(c). This
FIGURE 5.2 CVD system for CNT growth: (a) controller and furnace, (b) quartz tube in
furnace, (c) display of process temperature and flow rates, and (d) two samples on the loader
in the quartz tube.
module is a computer controller that regulates the furnace temperature and the
gas flow. The computer uses the Microsoft Windows environment and LABVIEW
as the programming language. The sample on the loader in the quartz tube is
shown in Fig. 5.2(d). The system uses four process gases: methane, ethylene,
hydrogen, and argon, which are involved in the synthesis of nanotubes. Nitrogen
is used for operating the pneumatic gas valves in the furnace. Methane and
ethylene are the carbon precursor gases. Hydrogen is used for moderating the
chemical reaction of the hydrocarbon decomposition. Argon creates an inert
atmosphere during the purging and cooling process steps. Safety is a concern
when using the EasyTube
TM
System and combustible gases. Our CVD system is
enclosed in a large fume hood with safety glass doors. In addition, leak detectors
for hydrogen and hydrocarbons are installed in the gas cabinet and all along the
gas supply lines to the furnace.
A gas supply system consisting of mass flow controllers and valves controls
the flow rate and the sequence of the gas flows. The gases enter the quartz tube
reactor where they are heated to a temperature of 9001000qC. The reactor pressure
is near atmospheric. The hydrocarbons decompose in the quartz tube over a catalyst,
which is mounted on a substrate. The catalysts used in our experiments include Fe,
Ni, and Co metals. Ceramic powdered materials such as alumina, silica, or magnesium
oxide support them during the CVD growth. A quartz or molybdenum boat is used
to accommodate a powder-supported catalyst for bulk nanotube growth. In some
of our experiments, a Si substrate patterned with catalyst is used for surface
synthesis of CNT.
In situ quadrupole mass spectroscopy (QMS) is a very useful technique for
monitoring the species in the CVD reactor, and the related studies can improve the
quality of the CNT. A dynamic mode of data acquisition can be employed to track
the concentration of the major species with time. This mode easily reveals the
changes in the partial pressure of the species when changing the concentration of
methane or ethylene by pulsing of the carbon precursor flow. The pulsing flow
approach is described later in this chapter. The acquired spectra usually illustrate
change in the concentration of the major species such as C
, H
2
CH
3
, and other
hydrocarbon radicals, which change with the carbon precursor pulses. This tech-
nique was successfully tested during plasma CVD growth of polycrystalline dia-
mond films.
33
To investigate the gas phase environment during the nanotube growth, the Easy-
Tube
TM
system was modified by incorporating a quadrupole mass spectrometer,
VISION 1000 P-M (300 amu), made by MKS Instruments. This instrument samples
from the reactor during the deposition and provides quick feedback information
about the species in the reaction zone. This approach helps optimize the process
variables in terms of quality of the CNT. The modified CVD system with the mass
spectrometer coupled to the reactor is illustrated in Fig. 5.3.
Sampling of the gas phase is done through a heated capillary inserted inside the
reactor. The capillary is connected to the QMS sensor via a pressure-reducing system,
allowing measurement at close to atmospheric pressure in the CVD reactor.
Pictures of the QMS incorporated into the CVD system are shown in Fig. 5.4.
The extension of the heated capillary through a stainless steel probe is illustrated in
Fig. 5.4(a). The capillary goes inside of the reactor and its inlet is close to the boat
with the sample. This arrangement allows sampling of the gas phase within the
vicinity of the growth zone. A 7 Pm filter built up along the line between the capillary
and the sensor protects the sensor from solid particles found in the reactor. Figure
5.4(b) shows the main components of the QMS.
FIGURE 5.3 Schematic of the CVD system with the incorporated QMS.
FIGURE 5.4 Pictures of the QMS incorporated into the CVD system: (a) heated capillary
introduced into the reactor through a stainless steel probe and (b) major modules of the QMS.
Ouartz
tube
Gas inlet
Sample
Capillary
QMS
sensor
Turbo pump
Dry diaphragm
pump
Atmosphere
Furnace
Quartz
boat
C
2
H
4
CH
4
H
2
Probe
Sensor
Dry pump
Controller
Capillary Turbo pump
(b) (a)
5.3 CATALYST AND SUBSTRATE PREPARATION
One of the approaches tested in this work was to perform bulk synthesis of CNT
using MgO powder as the supporting substrate for the metal catalyst. Our choice
for MgO is based on previous experience with Al
2
O
3
and SiO
2
as supporting ceramics
that are difficult to remove during purification of the CNT. Magnesium oxide allows
easy purification of CNT using hydrochloric acid. The substrate is prepared by
thermal decomposition of Mg(NO
3
)
2
or MgCO
3
. The preparation of the metal catalyst
on the MgO substrate is a multistep process. The ceramic powdered substrate is
immersed into a Fe(NO
3
)
3
aqueous or methanol solution and dispersed by high-
power ultrasound. Later it is separated from the solution and dried at 100qC for
2 h.
20
In our experiments the preceding procedure was modified as follows: The
ceramic powder was dispersed in water by high-power ultrasound. Then the dis-
persed powder was diluted by deionized water, and Fe(NO
3
)
3
6H
2
O was added to
the MgO suspension. The mixture was treated for 30 min in an ultrasonic bath for
further dispersion and dried at 100qC in air. The dried powder was ground in an agate
mortar to reduce the grain size. In some experiments, Bis(acetylacetonato)-dioxymo-
lybdenum (VI) is added to prepare a bimetal Fe-Mo catalyst.
34
As an alternative,
Mo salt such as (NH
4
)
6
Mo
7
O
24
4H
2
O can be used for the same purpose.
9
Further,
the ceramic substrate impregnated with catalyst compound is calcinated in air at
500qC for 1.5 h. This step results in the formation of metal catalyst oxides. The last
step in preparation of the catalyst takes place in the CVD reactor prior to the
deposition of CNT. There the metal oxides undergo reduction with hydrogen and
transform into metal particles, which cover the MgO powder.
Another approach we are exploring is the deposition of the catalysts on an
oxidized Si substrate, patterned by photolithography, as shown in Fig. 5.5. This
technique facilitates growth of high-purity nanotubes.
26
A suspension containing
a catalyst salt such as Fe(NO
3
)
3
and alumina particles in methanol was selectively
applied by spin deposition on the Si substrate, followed by baking at 160qC for
5 min. Spin coating provides close control of the liquid film and of the catalyst
thickness. It is important to note that for selective deposition of the catalyst, a
photoresist mask of polymethylmethacrylate (PMMA) was used. In Fig. 5.5, the
windows within the photoresist were filled with catalyst suspension. The windows
can be square, circular, or cross-shaped. Prior to the high temperature growth, the
photoresist was removed by a standard liftoff procedure using acetone. The substrate
was then thermally treated in the CVD reactor to decompose the metal salt and to
produce the metal catalyst by reduction with hydrogen. Alternatively E-beam lithog-
raphy can be used to create islands of metal catalyst sputtered on silicon wafers for
growth of aligned carbon nanotubes.
Photos of the catalysts, substrates, and substrate carriers for SWCNT growth
are shown in Fig. 5.6. The substrate holder in Fig. 5.6(a) is made of quartz and is
designed for surface synthesis of CNT. It allows carrying a Si substrate inside of
the CVD reactor, which can be patterned on both sides. When the patterned side is
facing down, the growth direction is the same as that of the gravitational forces.
Catalysts and substrates used in our experiments are shown in Fig. 5.6(b). For bulk
synthesis, the powdered catalyst is spread at the bottom of a Mo boat as a thin layer
in order to prevent any diffusion limitations during the growth of CNT. The Mo boat
with the catalyst is shown in Fig. 5.6(c).
5.4 GROWTH OF CNT
For CVD growth of CNT, the substrate with the catalyst is inserted automatically
inside the quartz reactor via a quartz loader tray driven by the computer. There are
many recipes developed and currently in use at UC for synthesis of CNT at different
conditions. The recipes are preprogrammed in a library on the computer that drives
the CVD system. New recipes can be easily added. The standard recipe is capable of
producing high-yield and high-purity SWCNT in three process steps: (1) maximum
ramp, (2) deposition, and (3) cooling. In the maximum ramp step, the temperature
of the furnace is raised from ambient to 900qC in 12 min. A continuous flow of
FIGURE 5.5 Patterned silicon substrate with catalyst before and after removal of the photoresist
mask: (a) plane view before removal of the photoresist, (b) cross-section before removal of the
photoresist, and (c) cross-section after removal of the photoresist.
(b)
(c)
Catalyst
Silicon substrate
Photoresist
mask
10 m 10 m
Catalyst
Silicon substrate
10 mm
(a)
Catalyst:
Ni or Fe or Co
supported by
Al
2
O
3
Photoresist
mask
argon is used to properly purge the reactor. Once the temperature reaches 900qC,
it is maintained for 10 min. This is the deposition stage. Nanotube synthesis occurs
during this stage when the supply of argon is stopped and the carbon source gases
methane and ethylene are run for 10 min along with a flow of hydrogen. After the
nanotubes are synthesized, they are cooled to ambient temperature, which completes
the last process step. During cooling the methane, ethylene, and hydrogen flow is
stopped, and the system is purged with argon. Apart from these three basic steps, a
separate step called the timed ramp can be incorporated. Unlike maximum ramp in
which the temperature rises as fast as possible, in the timed ramp the temperature is
programmed to rise gradually. The variables responsible for nanotube growth include
the temperature and duration of the process steps, the nature of the carbon precursor
and its flow rate, the type and quantity of the catalyst, and the selection of the
patterned substrate. These process variables determine the yield and quality of the
nanotubes synthesized. Manipulating any one of them can lead to changes in size,
shape, alignment, and the yield of the nanotubes. Many experiments are required
to determine the optimum process parameters for fabrication of CNT with desired
FIGURE 5.6 Photos of catalysts and substrates for CNT growth: (a) UC substrate holder for
surface synthesis, (b) catalysts and CNT, and (c) molybdenum boat with catalyst for bulk
synthesis.
Quartz
holder
Patterned
silicon
substrate
Patterned
silicon
substrate
(a)
(b)
(c)
Liquid
catalyst
Powdered
catalyst
Bulk
CNT
Uniformly scattered
catalyst
Heap of catalyst
Molybdenum boat
qualities. Table 5.1 lists a set of process conditions for a standard recipe used in
our experiments to grow SWCNT on MgO-supported Fe-Mo catalyst.
Another set of parameters designed and tested at UC is shown in Table 5.2. The
process parameters for Experiment 2 were selected based on optimization of the
CVD growth in terms of high yield and good quality of the CNT. The changes
compared to Experiment 1 include reduced gas flow rates and increased deposition
temperature and deposition time. These growth conditions produced good results,
which are presented later in the chapter.
The deposition conditions for Experiment 2 are graphically illustrated in Fig. 5.7.
The FirsNano CVD system allows on-line monitoring of the temperature and the gas
flow profiles while the experiment is in progress.
A third set of parameters tested at UC is displayed in Table 5.3. The growth
conditions for Experiment 3 were tuned to promote synthesis of MWCNT. Compared
to Experiments 1 and 2 the flow rate of ethylene is increased and methane is not
used. In addition, the hydrogen flow rate is reduced, which enhances the decompo-
sition of the hydrocarbon precursor. This causes an increased carbon flux moving
toward the substrate in the CVD reactor, which results in higher supersaturation of
the carbon atoms within the growth zone, and especially in the metal catalyst-carbon
solution. The overall effect is growth of MWCNT. Good-quality and high-yield
MWCNT can be produced following the recipe in Table 5.3.
TABLE 5.1
Standard Recipe for Growth of SWCNT, Experiment 1
Step Type
Temperature (qC)
Time (min)
Flow Rate (SCCM)
Start End Hydrogen Methane Ethylene Argon
1 Heat Ambient 900 12 0 0 0 1000
2 Deposition 900 900 10 300 1200 30 0
3 Cool 900 Ambient 120 0 0 0 1000
d
TABLE 5.2
New Recipe for Growth of SWCNT, Experiment 2
Step Type
Temperature (qC)
Time
(min)
Flow Rate (SCCM)
1 Purge Ambient Ambient 10 0 0 0 1000
2 Timed Ramp Ambient 500 10 0 0 0 1000
3 Purge 500 500 20 0 0 0 1000
4 Maximum Ramp 500 900 8 0 0 0 1000
5 Deposition 900 900 60 100 400 10 0
6 Cool 900 Ambient 120 0 0 0 1000
For bulk synthesis the catalyst is spread as a thin layer on the bottom of a
molybdenum boat. The scanning electron microscope (SEM) photos of the CNT
synthesized during this and other experiments are shown later in the characterization
section of this chapter. The MgO supporting ceramic is easy to remove by purification
FIGURE 5.7 Graphical interface of operations for Experiment 2.
TABLE 5.3
New Recipe for Growth of MWCNT, Experiment 3
Step Type
Temperature (qC)
Time
(min)
Flow Rate (SCCM)
1 Purge Ambient Ambient 5 0 0 0 1000
2 Timed Ramp Ambient 400 10 0 0 0 1000
3 Purge 400 400 5 0 0 0 1000
4 Maximum Ramp 400 700 8 0 0 0 1000
5 Deposition 700 700 30 50 0 250 0
6 Cool 700 Ambient 120 0 0 0 1000
Temperature profile
Flow profile
1000
800
600
400
200
0
1000
800
600
400
200
0
0 20 40 60 80 100 120 140 160 180 200 228
0 20 40 60 80 100 120 140 160 180 200 228
T
e
m
p
e
r
a
t
u
r
e

(
C
)
F
l
o
w

(
s
c
c
m
)
Time (min)
Time (min)
Hydrogen
Methane
Ethylene
Argon
with acids. Different combinations of MgO-supported Fe-Mo catalyst are being
tested to optimize the growth. With the long molybdenum boat and a thin coating
of the substrate-catalyst spread on the bottom, approximately one-half gram of
SWCNT can be produced in each experiment using our CVD system. The total
processing time for our standard synthesis experiment is about 3 h. The purification
of CNT, which includes acid treatment, thermal annealing, and washing, takes a
couple of additional hours. Running simultaneously, synthesis and purification of
different batches can produce 12 g of SWCNT per day. The use of more catalyst,
a larger boat, and an appropriately modified CVD reactor can all increase the yield.
New growth experiments are currently underway at UC that are expected to reach
this objective. Other benefits may also be possible, such as increasing CNT length,
and controlling the chirality and the alignment of the CNT.
Pulsed carbon precursor delivery during CVD growth of CNT is currently being
developed at UC. It is known that high-quality CVD diamond films can be achieved
by controlled manipulation of the growth parameters and particularly by changing
the carbon precursor flow during the deposition process. Time-modulated CVD
(TMCVD) is reported to produce smoother, harder, and better quality diamond
films.
35,36
The results published in these papers show controlled diamond film growth
by manipulating the flow of the carbon precursor. Another approach that is very
similar to TMCVD is a two-step growth process. This is especially useful for low-
temperature deposition, where it is difficult to obtain smooth films.
37
This process
promotes secondary nucleation, which changes the product morphology. Recently,
V. Shanov et al. reported pulsing of methane flow between a high and low level
during the CVD process for producing small-grain-size diamond films.
33,38,39
The
high methane concentration flow increased the carbon supersaturation on the sub-
strate and enhanced secondary nucleation. This results in small grain size of the
deposit, but a higher nondiamond (sp
2
) phase in the film.
This previous experience is employed in pulsing of the carbon precursor delivery
during the CVD process for better control of the quality and quantity of the CNT.
For this purpose, the EasyTube
TM
furnace was computer-programmed to pulse the
methane flow in the CVD reactor. This approach offers the experimental opportunity
to manipulate the carbon supersaturation in the catalyst-carbon solution during the
growth of the CNT. Another outcome of this process is expected to be the manu-
facturing of longer CNT grown in a vertical reactor. The study of the gas environment
within the CNT growth zone is done based on measurements from the quadrupole
mass spectrometer (QMS). The QMS allows in situ identification of the characteristic
ionic and molecular species in the CVD reactor.
The magnesium oxidesupported iron catalyst for bulk growth, along with the
patterned Si substrate for surface synthesis, described in Reference 40, was
employed for these experiments.
40
Mixtures of CH
4
and H
2
are used as reactant
gases for the CVD growth. The growth conditions for two experiments are shown
in Table 5.4. These experiments were performed in two modes: (1) the continuous
mode and (2) the pulsed mode. In the continuous mode the methane flow is
maintained constant throughout the deposition process. During the pulsed mode,
the methane flow is pulsed between a high and a low level. This procedure is
automated by employing computer-controlled mass flow controllers. The duty
cycle of the modulated methane flow is selected to be 50%, and the total time of
one cycle is 10 min. This is illustrated in Fig. 5.8. The deposition time is maintained
at 15 min for the nonpulsed flow experiments and is doubled for the pulsed flow
experiments. During the growth step the hydrogen flow is maintained constant,
but in future studies it will be manipulated along with the carbon precursor flow,
and also with the floating catalyst flow.
The dynamic mode of the QMS was used to track the concentration of the major
gaseous species with time. This mode conveniently reveals the change in the partial
pressure of the species when the methane gas flow is pulsed from high to low and
so on. The acquired spectra for continuous and pulse mode of the methane delivery
are shown in Fig. 5.9. The four major species monitored in the CVD reactor during
the CNT growth are H
2
(AMU 2), C
(AMU 12), CH
3
(AMU 15), and CH

4
(AMU 16). The partial pressure of CH

3
is close to that of CH
4
, and both species

follow the same chart trend. We selected to reveal in Fig. 5.9 only the CH
4
species
TABLE 5.4
Deposition Conditions for Continuous and Pulsed CVD Growth
of CNT
Growth Time
(min)
Temp.
(qC)
Hydrogen Flow
(sccm)
Methane Flow
Mode
Methane Flow
(sccm)
15 950 100 CM
*
600 (high)
30 950 100 PM
**
600-0 (high-low)
*
Continuous mode.
**
Pulsed mode.
FIGURE 5.8 Modulated carbon precursor flow for growing of CNT.
Low
High
5 min
10 min
5 min
M
e
t
h
a
n
e

a
n
d
/
o
r

e
t
h
y
l
e
n
e
f
l
o
w

(
s
c
c
m
)
Time (minutes)
along with H
2
and C
. During the first growth experiment the methane flow rate

was kept constant and the obtained QMS spectra are shown in Fig. 5.9(a). The
concentrations of the three major species remain constant, which is an expected
result. The amount of CH
4
dominates and C
+
reveals the lowest partial pressure.
Figure 5.9(b) shows the results from the second experiment where methane flow
pulses from 600 sccm to 0 and back to 600 sccm. The observed concentration of
CH
3
and C
species first decreases and then increases accordingly. Because of

residual methane in the reactor during the off period, the partial pressure of carbon-
containing species does not reach zero value. The flow rate of hydrogen remains
constant during the entire deposition time, but the partial pressure of H
2
+
increases
with turning the methane flow off and vice versa.
The experiments conducted confirmed that the spectra of the monitored species
are very sensitive to any flow changes and provide important information about the
gas phase environment in the CNT growth zone.
These may be the first results reported using pulsed flow and residual gas
analysis for CNT growth. Further studies are under way with more in-depth
analysis of the CVD growth. Different duty cycles will be explored to refine the
proposed pulsed carbon precursor flow technique. The obtained data will be
correlated with Raman spectroscopy, SEM, and transmission electron microscope
(TEM) studies of the CNT. Our preliminary results indicate that pulsed CVD with
residual gas analysis will provide better control of the purity and morphology of
the CNT in terms of diameter, growth alignment, and number of walls. This novel
approach is more complex than standard CVD of CNT, and further research is
needed to develop it.
FIGURE 5.9 QMS spectra of CNT growth run in two different modes of methane delivery
at conditions shown in Table 5.4: (a) continuous methane flow, and (b) pulsed methane flow.
1000
900
800
700
600
500
400
300
200
100
1000
900
800
700
600
500
400
300
200
100
0 5 10 15 20 25 30
0 5 10 15 20 25 30
Methane ON
Methane
ON
Methane
OFF
CH
4
+
H
2
+
+
C
CH
4
+
H
2
+
+
C
P
r
e
s
s
u
r
e

(
t
o
r
r
)
P
r
e
s
s
u
r
e

(
t
o
r
r
)
Time (minutes)
Time (minutes)
(a) (b)
5.5 PURIFICATION OF AS-GROWN CNT
As deposited CNT are frequently contaminated with metal catalysts, ceramic
particles, amorphous carbon, and graphitic nanoparticles, purification of the as-
grown CNT is a complex procedure. The first step of purification includes removal
of the impurities by wet chemical oxidation. Hydrochloric and nitric acids are
frequently used for this purpose. Additional steps include decanting, centrifuga-
tion, or filtration, followed by rinsing in deionized water. Gas or vapor phase
oxidation is also used for deeper refinement. Finally gentle treatment in an ultra-
sonic bath helps to continuously break up agglomerated nanotubes. These
approaches are frequently combined to achieve more comprehensive purification
and a higher CNT yield.
Alumina/silica as a powder substrate was used in our initial experiments. These
two oxides are hard to dissolve chemically, and HF acid was employed for removal
of silica. This acid is difficult to handle due to its hazardous nature. Magnesium
oxide
19,20
is easier to dissolve and is now routinely employed in our current research.
The purification procedure involves sonication of the unpurified CNT sample in
HCl acid with a concentration of 4 M for 1 h, and additional soaking of the sample
in acid for 24 h. Washing in water and filtering of the CNT follows these steps.
A simple approach for purification of MWCNT is high temperature oxidation
(325425qC) in air. It is important to note that when using this technique the metal
catalyst should be removed first by wet oxidation. The last step is required, since
in the presence of oxygen the metal particles catalyze low-temperature oxidation of
the CNT, which decreases the yield dramatically.
41
Purity evaluation of the CNT is
also not a simple procedure. The metal catalyst weight in as-grown CNT is about
30%. Measuring this contaminant can be done by thermogravimetric analysis (TGA).
Burning the CNT sample in air at 1000qC removes all the carbonaceous material
and the TGA measurement gives the metal catalyst content in the original sample
converted to oxides. Measuring the carbonaceous purity is a much more difficult
procedure. Most of the qualitative and quantitative techniques have been based on
spectroscopy. The ideal technique has to distinguish between graphitic and amor-
phous carbon phases in the CNT. A very promising approach for this is solution-
phase near-infrared (NIR) spectroscopy. More details about this technique can be
found in Reference 41.
Dry methods for purification of CNT represent another attractive approach, since
they offer minimum loss during the treatment. Dry purification requires plasma and
is more expensive than wet etching because of the cost of the plasma facility. Plasma
etching for removal of amorphous carbon that covers aligned CNT has been tested
using water plasma.
42
In addition, this treatment removes the end caps of the metal
catalysts from the top of the CNT. However, a prolonged plasma procedure may
disintegrate the graphitic structure of the nanotubes. Dry etching with H
2
O and with
H
2
plasma is a promising area that should receive further experimental investigation.
Additional information on purification of CNT synthesized by different methods is
provided in References 4346.
5.6 CHARACTERIZATION OF CNT
Carbon nanotubes are extremely small objects and their characterization requires
sophisticated instruments. The morphology of the CNT, their dimensions, and ori-
entation can be easily revealed using scanning electron microscopes with high
resolution.
6,10,17,18,2326
Appropriate is the use of an environmental SEM (ESEM),
which does not need preparation of the samples by conductive coating. Details on
ESEM and its use for our research are given in a later section. Another advanced
approach is atomic force microscopy (AFM), which provides the opportunity to
study features and properties of CNT that are not available with other techniques.
10
High-resolution transmission electron microscopy (HRTEM) is widely used to char-
acterize the overall shape and diameter of the CNT, the number of the walls, their
thickness, and the distance between the walls.
8,9,15,17,18,20,47
In addition, the electron
diffraction mode of the TEM helps to identify the nature of the cap on top of the
CNT, which is usually composed of the metal catalyst.
14,23,24
Micro-Raman spectros-
copy is frequently employed to study the quality of the carbon nanotubes.
20,21,27,48
The tube configuration can be investigated in detail using this technique.
8
It also
provides information about the number of the walls and the presence of crystalline
or amorphous carbon, including the diameter of SWCNT.
5.6.1 SCANNING ELECTRON MICROSCOPY
A FEI XL-30 ESEM FEG made by Philips has been used to study the morphology
of the CNT. A Shottky hot-field emission tip is employed as an electron source and
has an ultimate resolution of 1.21.5 nm. A large specimen chamber housing a
motorized stage with an internal CCD camera allows observation of fairly large
specimens. Thanks to the variable pressure limiting apertures, the ESEM can be
operated at a pressure chamber of 1 to 20 torr. Gaseous detection systems are used
in imaging the samples. Partial ionization of the chamber gas causes charge neu-
tralization at the sample surface, with the oppositely charged species being collected
by the gaseous secondary detector through a cascade effect. The overall outcome is
an improved image of samples regardless of the nature of the studied samples. Some
of the samples discussed were observed using a Hitachi S4000 SEM.
Two groups of CNT samples grown at UC are studied by ESEM. The first group
includes specimens synthesized on a flat Si substrate, and the second represents bulk
synthesis of CNT on powdered ceramic carriers. The samples are directly mounted
onto a standard aluminum fixture using double-sided carbon adhesive tape. No further
sample preparation or metallization is carried out, allowing the nanotubes to be
imaged in their natural condition. The following set of parameters is used to obtain
high-resolution images of CNT: working distance of 810 mm, accelerating voltage
of 1030 KV, and a chamber pressure of 0.91.3 torr. The images in Fig. 5.10 were
taken using the ESEM. They illustrate unpurified nanotubes grown in Experiment
1 on alumina particles coated with the Fe-Mo catalyst patterned on a Si substrate.
The images reveal CNT dimensions of about 1 Pm long and 23 nm in diameter,
which suggests that the specimens are SWCNT. The nanotubes form a spaghetti-
type pattern with tubes lying over each other. Figure 5.10(a) displays typical
morphology of SWCNT grown on a catalyst island. Excessive shining catalyst is
present below the nanotubes, which can be removed by purification with acids.
Figure 5.10(b) shows growth of SWCNT between catalyst particles on a silicon
substrate, indicating that CNT could be synthesized by bridging catalyst particles.
It is observed that long SWCNT originate from one mound of catalyst and terminate
at a different mound of catalyst running over the dark silicon substrate. This suggests
the possibility for direct growth of a nanotube network between patterned islands
on the substrate in two dimensions.
The images in Fig. 5.11 were taken using the Hitachi S4000 SEM. Figure 5.11(a)
shows purified CNT filaments grown in the modified Experiment 2 using a powdered
MgO substrate coated with Fe catalyst. In this experiment ethylene was not present
and the methane gas flow rate was increased up to 420 SCCM to compensate for
the absence of ethylene precursor. The growth time was 30 min. Figure 5.11(a)
shows CNT filaments as well as isolated structures with smaller diameter. The
samples have been further investigated by TEM and the results are presented later
in this chapter. Figure 5.11(b) shows MWCNT grown in Experiment 3 using a
powdered MgO substrate coated with Fe catalyst. The spaghetti-like bundles of
MWCNT are further studied and reported later in the TEM section. The morphology
and the dimensions of the CNT are very sensitive to the growth conditions and may
change after purification.
17,18,42
Figure 5.12 illustrates bulk nanoropes grown on powdered MgO substrate
coated with Fe catalyst. The images contain bundles of CNT ropes with diameters
of 4080 nm. Further study of these structures was conducted by TEM and the
results are reported in the next section.
The bundle-type morphology shown in Fig. 5.11 and Fig. 5.12 is similar to that
obtained by other groups
17,18,34
and proves that bulk synthesis, although good for
high-yield manufacturing, cannot provide aligned growth.
FIGURE 5.10 SEM images of purified SWCNT grown in Experiment 1 using alumina
supported Fe-Mo catalyst on a patterned silicon substrate: (a) SWCNT at magnification of
u200,000, and (b) SWCNT grown between catalyst particles at magnification of u50,000.
5.6.2 TRANSMISSION ELECTRON MICROSCOPY
In transmission electron microscopy (TEM), a thin solid specimen (200 nm thick)
is bombarded in vacuum with a highly focused, monoenergetic beam of electrons.
The beam is of sufcient energy to propagate through the specimen. A series of
electromagnetic lenses then magnies this transmitted electron signal. Diffracted
electrons are observed in the form of a diffraction pattern beneath the specimen.
This information is used to determine the atomic structure of the material in the
sample. Transmitted electrons form images from small regions of the sample that
FIGURE 5.11 SEM images of puried CNT synthesized by bulk growth on MgO supported
Fe catalyst: (a) CNT laments grown in modied Experiment 2, magnication of 100,000,
and (b) MWCNT grown in Experiment 3, magnication of 100,000.
FIGURE 5.12 SEM images of bulk puried nanoropes grown on powdered MgO substrate
coated with Fe catalyst at conditions of Experiment 2: (a) nanotubes in ropes, 10,000, and
(b) nanoropes at higher magnication, 25,000.
contain contrast, due to several scattering mechanisms associated with interactions
between electrons and the atomic constituents of the sample. Analysis of transmitted
electron images yields information both about the atomic structure and defects
present in the material. The lateral resolution is better than 0.2 nm on some
instruments.
49
A JEOL JEM 4000EX TEM instrument was used to acquire TEM images of
CNT. The CNT samples are dispersed in water using a bath sonicator for 30 min.
Two drops from the dispersed sample are dropped onto a supporting nickel grid and
dried in air. The microscope is operated at 400 kV.
Figure 5.13 shows high-resolution TEM images of purified bulk SWCNT syn-
thesized from MgO powder with Fe catalyst. The growth process was based on a
modified Experiment 1, which was run at a longer deposition time of 60 min. The
estimation of diameter of the SWCNT from the images shown in Fig. 5.13 is about
3 nm. The outer surface of the nanotubes looks rough which might be caused by an
amorphous carbon coating. The high methane flow rate of 1200 sccm and longer
deposition time may be the reason for this coating.
Figure 5.14 displays low-resolution TEM images of purified CNT filaments and
bulk MWCNT synthesized on a MgO-supported Fe catalyst. Figure 5.14(a) shows
CNT filaments grown in a modified Experiment 2, as described earlier. The filaments
consist of bundles of CNT, which might consist of SWCNT. Analogous images of
bundles of SWCNT are shown in References 8 and 10. Figure 5.14(b) exhibits
MWCNT grown in Experiment 3. It reveals a typical bamboo-like structure of
MWCNT with diameters ranging from 1530 nm. Similar morphology is described
in References 13, 17, and 18.
TEM images of nanoropes are shown in Fig. 5.15. The average diameter of the
ropes is in the range of 4060 nm. Figure 5.15(a) shows ropes bent in a U-shape.
Figure 5.15(b) shows a TEM image of a CNT nanocoil. Similar carbon nanocoils
with a 3-D spiral structure grown on a tungsten substrate using a nickel catalyst are
described in Reference 50.
FIGURE 5.13 Purified SWCNT grown in modified Experiment 1 from MgO powder with
Fe catalyst: (a) TEM image of SWCNT, and (b) TEM image of SWCNT at 2 times lower
magnification than that used in Fig. 5.13(a).
5.6.3 RAMAN SPECTROSCOPY
The Raman technique is used in qualitative and quantitative analysis of carbon
nanotubes. When a beam of light traverses a sample of a chemical compound, a
small fraction of the light emerges in directions other than that of the incoming
beam. Most of this scattered light is of unchanged wavelength. However, a small
part has wavelengths different from those of the incident light, and their presence
is a result of the Raman effect. The pattern of the Raman spectrum is characteristic
of the particular molecular species, and its intensity is proportional to the number
FIGURE 5.14 Low-resolution TEM images of purified CNT synthesized by bulk growth on
MgO-supported Fe catalyst: (a) CNT filaments grown in Experiment 2, and (b) MWCNT
grown in Experiment 3.
FIGURE 5.15 High-resolution TEM images of MWCNT grown at UC: (a) low-magnification
image of dispersed CNT, (b) high-magnification image of a MWCNT separated from the
array. Growth conditions: 200 sccm of H
2
, 200 sccm of C
2
H
4
, deposition temperature of
750C, and Fe catalyst patterned on oxidized Si substrate.
of scattering molecules in the path of the light. Resonance peaks are also observed
in the spectrum, which symbolize the presence of a particular species type that is
in abundance. Comprehensive information on the Raman technique and its applica-
tion can be found in References 5153.
Micro-Raman spectroscopy is used for characterization of CNT with a typical
excitation wavelength of 632.8 nm (He-Ne laser) or of 514.5 nm (Ar ion laser). The
diameter of single-wall CNT can be determined by SEM/TEM measurements and
additionally verified by micro-Raman spectroscopy.
The characteristic spectrum of SWCNT exhibits three main zones at low (100
250 cm
1
), intermediate (3001300 cm
1
), and high (15001600 cm
1
) frequencies.
8,20
These bands can be reduced by the presence of other carbon forms like multiwall
nanotubes or amorphous carbon.
20,27
Other carbon materials produce other types of
spectra.
54
The high-frequency bands can be decomposed in one main peak around 1580
cm
1
, with a shoulder around 1540 cm
1
. This shoulder is more important in the case
of SWCNT produced by arc discharge, where a separate peak can be observed.
20,27
In carbon-based materials, typically two main first-order peaks are present. One is
the D peak, observed around 1300 cm
1
(for excitation with He-Ne laser) or at 1350
cm
1
when using an Ar ion laser. The D peak is related to the presence of defects.
21
The other one is the G peak at about 1580 cm
1
, which is associated with the in-plane
vibrations of the graphene sheet.
48,55
In addition to the G line, the appearance of a
side peak at about 1540 cm
1
indicates the existence of single-wall nanotubes with
different diameters.
27,56
Ratios of the D peak to the G peak have been used as an
indicator of the amount of disorder within nanotubes.
21,57
A small I
D
/I
G
ratio, typically
in the range of 0.10.2, indicates that the defect level in the atomic carbon structure
is low, which suggests reasonable crystalline quality.
48
The high-resolution spectrum obtained in the low-frequency domain shows
several components within the range of 100250 cm
1
.
20,21
Prominent low-energy
peaks around 191 and 216 cm
1
are the breathing modes of nanotubes vibrating
along the radial direction
56
and can be clearly revealed by using a He-Ne laser.
27
The spectrum in the low-frequency domain reflects the SWCNT diameters and can
be used to evaluate it. In general, the frequency increases with decreasing tube
diameter ( d ). The frequency v of these modes is known to be inversely proportional
to the diameter of the SWCNT. Based on the equation v (cm
1
) 223.75/d (nm)
the diameter of the SWCNT can be determined.
58
The calculated ( d ) values reflect
a distribution in the diameters of the CNT.
The peaks ranging from 400 to 1000 cm
1
are usually observed in single-wall
nanotubes when investigated using as excitation wavelength of 632.8 nm and could
be related to the finite length of the carbon nanotubes.
27,59
In the case of MWCNT studied using excitation with a He-Ne laser, the G
line reveals a small bump at the higher energy side around 1620 cm
1
and it is
attributed to the multiwall structure. In addition, breathing modes are significantly
diminished.
27
The Raman spectrum of nanotubes grown by thermal CVD on Co-
coated Si substrates shows the G-line peak at 1587 cm
1
.
27
The observed broad peak
near 1337 cm
1
indicates the existence of defective graphitic layers on the wall
surfaces due to the relatively low growth temperature. Prominent peaks on the lower
energy side cannot be observed in this case.
27
Another study claims investigation of
low-frequency spectra of multiwalled CNT.
60
Low-frequency modes between 115
and 220 cm
1
are obtained after purification of the carbon nanotubes and excitation
with a red laser line of 676 nm and with a green line of 514 nm. These radial-
breathing modes originating from individual walls can be used for characterization
of multiwalled CNT.
60
The excitation energy of the laser sources can be used to explore the spectrum in
the range 14001700 (cm
1
) and find variations in the metallic or semiconducting
nature of the fabricated CNT. If the light excitation energy is between 1.7 and 1.9 eV,
and resonance is observed in the 14001700 cm
1
wave number range, it is an
indication of metallic nanotubes. If the light excitation energy is above 2.2 eV or
below 1.5 eV, and resonance peaks are observed in the 14001700 cm
1
wave number
range, it is evidence of the semiconducting nature of the CNT.
61
Details on Raman spectroscopy of CNT are presented in Chapter 3 of this book.
An inVia Raman microscope was employed to obtain the CNT spectra synthesized
in three runs based on Experiment 1. The Renishaw spectrometer is a 250-mm focal
length system. For the 514.5 nm laser excitation, 1800 l/mm grating, and 50 Pm slit
width, the spectral resolution is about 5.5 cm
1
. This spectrometer is coupled with a
RenCam high-performance cooled CCD detector, which is an all-in-one compact unit.
It uses thermoelectric cooling to operate at 70qC. Its detector chip and electronics are
carefully engineered to offer users ultra-low noise and low dark current.
One of the Raman spectra obtained in Experiment 1 is shown in Fig. 5.16. In this
analysis only wave numbers 0300 (cm
1
) are considered for calculation of diameter
of SWCNT, and the results are presented in Table 5.5. The relationship between
diameter and resonance wave number is v (cm
1
) 223.75/d (nm). The CNT diameters
are almost the same in the three experiments and are in the range of 1.21.3 nm.
FIGURE 5.16 Raman spectrum of CNT grown in Experiment 1 from alumina-supported Fe-
Mo catalyst on a patterned silicon substrate.
1000
900
800
700
600
500
400
300
200
100
0
0 50 100 150 200 250 300 350
Wave number
I
n
t
e
n
s
i
t
y

(
c
o
u
n
t
s
/
s
e
c
)
5.7 ADVANCED TOPICS AND FUTURE DIRECTIONS
FOR CVD OF CNT
Our future work in the field of CVD synthesis of CNT includes three major directions:
1. The growth of long nanotubes is our long-term goal, which will have a
positive impact on development of a new generation of composite mate-
rials and sensors. Increasing the length of nanotubes will improve the
load transfer from the nanotubes to the matrix and thus the overall
mechanical properties of CNT-based composites. This objective will be
achieved by using our newly designed vertical CVD reactor, which will
be fed with floating catalysts. In situ monitoring of the gas environment
in the reactor by a quadrupole mass spectrometer is expected to provide
better control of the process parameters. Pulsed carbon precursor delivery
during the CVD deposition is currently being studied at UC, and the
preliminary experiments show improved quality control of the grown
CNT. These efforts are expected to enhance our abilities to synthesize
longer CNT.
2. Controlled alignment of SWCNT, selectively grown on flat substrates, is
another objective of our future research. This will open opportunities to
create a building brick for CNT sensors and actuators. An array based
on SWCNT for actuator applications has been designed and will be
manufactured in the near future at UC. Semiconductor processing tech-
niques will be used to fabricate the CNT array and devices.
3. Functionalization of the carbon nanotubes is another goal of our future
research. It will help us change the nanotube surface properties by apply-
ing a very thin layer of functional materials or attaching selected func-
tional groups. Different deposition techniques, including plasma CVD,
are among the options we are currently exploring to achieve this objective.
The controlled functionalization has a great potential for building
advanced sensors and actuators with applications in new areas, including
cancer diagnostics and therapy.
TABLE 5.5
Numerical Data Estimated from the Raman Spectra of SWCNT
Grown on a Silicon Substrate
Experiment Number
Resonance Wave
No. (cm
1
)
Diameter
(nm)
Experiment 1 192.7 1.2
Modified Exp. 1 (longer purging at elevated
temperature)
183.9 1.2
Modified Exp. 1 (longer purging at elevated
temperature and longer deposition time)
169.8 1.3
5.7.1 SYNTHESIS OF LONG CARBON NANOTUBES
An innovative aspect of nanomaterials synthesis involves the preparation of long
carbon nanotubes in a vertical CVD reactor.
29,30
In its current horizontal configura-
tion, the CVD system at UC can be used to make SWCNT that have a length of
about several micrometers. We are modifying the horizontal furnace to a vertical
configuration that it is hoped will produce SWCNT or strands with lengths of
millimeters or even centimeters. The vertical furnace will be equipped with instru-
ments to monitor the growth environment of SWCNT and to improve process control.
The vertical CVD reactor will be fed with a floating catalyst. A schematic drawing
of the CVD system with a vertical furnace is shown in Fig. 5.17. The vertical CVD
system will have several unique features, including fiber optic control of temperature
and individual preheat of all gases for fast growth. In addition, feedback from the
FIGURE 5.17 CVD system with a vertical furnace.
Catalyst
Adjustable
Feedstock
H
2
Reconfigurable
nanofurnace
for continuous
growth of
nanotubes and
strands
Synthesis
furnace
Reactor
CNT strand
Detachable collector
Exhaust
Fiber optic
pyrometer
Quadrupole mass
spectrometer tube
Adjustable
Rotatable
stand
Zone 1
preheat
Zone 2
synthesis
QMS to the mass flow controllers will provide precise control of the carbon and
metal-catalyst concentration in the growth zone.
Floating catalysts will be introduced in the described CVD reactor for gener-
ating nanometer-size particles of the metal-catalyst from the gas phase. The advan-
tages of this approach in a vertical CVD reactor include an uninterrupted downward
growth environment and alignment of nanotubes in the gas flow direction. In
addition, pulsed delivery of the carbon precursor will provide better control of the
carbon supersaturation in the catalyst-carbon solution. This is critical for the CNT
morphology and related properties.
5.7.2 DESIGN AND FABRICATION OF A SEGMENT
FOR CNT ACTUATOR
A promising approach to making nanotube sensors and actuators is the fabrication
of a single segment CNT array.
22,62
At UC we designed such a segment based on
fabrication steps that are frequently used in semiconductor processing technology.
Figure 5.18 illustrates the sequential steps in the fabrication process, which will
result in producing a building brick for CNT actuators. Two types of segments
are selected, depending on the electrolyte used for the actuator. The first one will
FIGURE 5.18 (Color figure follows p. 12.) Schematic illustration of the fabrication process
of CNT segment for actuator application.
Step 7
Metallization with Au or Ag
Au
Step 8
Step 9
Step 10
Deposition of doped poly Si
Sputtering of Co
Step 1
Step 2
Doped Poly Si
Step 3
Step 4
Step 5
Co
CNT
Step 6
Co or Ni
Deposition of Solid Electrolyte
Solid Electrolyte
High Resistivity Si Wafer or sapphire
Etching of SiO
2
or polymer
Electrical Wiring
SiO
2
Thermal Treatment At 825C in H
2
CoSi
x
CNT Growth
Deposition of SiO
2
or Polymer (EPON)
SiO
2
or
Polymer
perform in a liquid electrolyte. The second segment type will be embedded in a solid
electrolyte, which requires an additional step of processing, as revealed in Step 10 of
Fig. 5.18. The starting substrate (Step 1) could be either a highly resistive Si single-
crystal wafer or sapphire. This will provide an insulating base, which is required for
electrical wiring of the segment. Step 2 involves deposition of highly conductive
(doped) poly-Si, on which the CNT catalyst is planted. Doped Si will provide good
electrical contact between the CNT and the substrate. CVD technique will be
employed for depositing of poly-Si doped either with P or As. An alternative solution
is to use an insulating Si wafer, which can be ion implanted for increasing the
electrical conductivity. Steps 3 to 4 require sputtering of a thin metal film (2040
nm) of catalyst such as Co and thermal annealing of the film in hydrogen at 825qC.
The procedure is described in Reference 24. During the annealing, catalytic semi-
spheres of Co islands are formed, driven by surface tension to lower the total energy.
Simultaneously, a chemical reaction between the metal and the Si substrate takes
place, producing highly conductive CoSi
x
. This compound is expected to provide a
strong bond between the CNT and the substrate and will also serve as a conductive
electrical junction. The patterned substrate with catalyst provides aligned growth of
MWCNT during Step 4. Metallization of the top portion of the CNT for good
electrical contact requires several procedures. Step 6 fills up the gaps between the
individual CNT within the block with either SiO
2
or a polymer, which can be done
by standard deposition procedures for these materials.
22,62,63
This will give good
mechanical support to the CNT for the next metallization. Step 7 illustrates metal-
lization, performed by sputtering of Au on top of the CNT. Alternative metallization
is possible by brushing the top surface with silver paste.
Step 8 removes the supporting SiO
2
material by etching that will liberate the
CNT surfaces. Electrical wiring of the segment is shown in Step 9. Electrical signal
can be provided to each individual block/column of aligned carbon nanotubes using
the bottom wire.
In a new approach developed at UC, the top wire will be used to sense the strain
in the nanotubes. Since the resistance of the nanotubes will change with strain, the
array can become a self-sensing actuator. This will allow the actuator to have strain
feedback and become smart. The strain feedback will allow precise feedback control
of the actuator motion. More details on CNT actuators can be found in Chapter 15
of this book. Step 10 reveals the fabrication process for the second segment type.
The CNT blocks need to be embedded in a solid electrolyte. Steps 1 to 9 are identical
for both segment types and an additional Step 10 is added to the processing sequence.
This step will provide the solid electrolyte, which will embed the CNT columns.
The solid electrolyte will be deposited by plasma polymerization as discussed in
Chapter 8 of this book and Reference 64. The array growth shown in Fig. 5.18 is
feasible with our CVD system. Vertical MWCNT arrays grown on Si at the Universty
of Cinncinati are shown in Fig. 5.19.
5.7.3 ORIENTED GROWTH OF A SIMPLE CNT NETWORK
CVD is the most efficient method for synthesis of SWCNT with controlled orientation.
Different approaches have been used to enhance CVD-oriented growth, including exter-
nal gravitational, magnetic, and electric field. Electric fielddirected growth has been
successfully used to deposit aligned SWCN showing two-dimensional network fea-
tures.
65
Directed growth of free-standing SWCN on microtowers was demonstrated as
well.
66
Results have been published about growth of millimeter-long SWCNT, and also
about a two-dimensional nanotube network grown by a fast heating CVD process.
67
The controlled orientation during the growth process and deposition of a CNT network
is extremely important for developing artificial neural microsystems.
Our future research objective is to synthesize a simple network of parallel SWCNT
with a preset orientation, which is suitable for incorporation into a material neural
system for sensing of microcracks, defects, and corrosion. The approach is based on
two strategies: (1) to use a vertical CVD reactor for growth of long CNT, which has
been described previously, and (2) to perform growth experiments where lateral electric
fields will be applied across the prepatterned metal catalyst Si substrate during the
CVD process. Different electric fields such as DC, pulsed DC, or AC field will be
tested. The electrodes can be made of doped Si or other conductive materials. Align-
ment guidance by the field is expected to promote growth of the CNT network between
parallel electrodes or between microposts on the Si surface. This concept is illustrated
in Fig. 5.20. In addition, gravitational force experiments will be conducted to enhance
oriented growth of CNT. A substrate holder for such experiments, shown in Fig. 5.6(a),
has been made and is currently being tested. It will allow growth of the CNT in an
upside-down direction, taking advantage of gravitational forces.
5.7.4 FUNCTIONALIZATION OF CNT
Functionalization of CNT is essential in changing their overall properties. It can be
done by specific surface modification such as deposition of thin coating or attachment
of selected functional groups. Different coatings can be tailored to alter the electrical
and mechanical properties of the CNT due to a combination of dimensional changes
and interfacial properties.
Low-temperature functionalization of CNT is described in Reference 68. A micro-
wave plasma source is used in this study to generate hydrogen plasma, which is passed
FIGURE 5.19 Long-array nanotubes grown at the University of Cincinnati: (a) a 4-mm
MWCNT patterned array; and (b) SEM side view of a nanotube array at 1,000u magnification.
The nanotubes were grown by Yun Yeo-Heung. The patterned substrate was provided by Yi Tu
of First Nano Inc.
through a curved tube into the process chamber. The atomic hydrogen species reach
the CNT and cause their hydrogenation. The process is dry, clean, and relatively simple.
Plasma polymerization coating of nanotubes is a novel approach used at UC to
improve the surface properties of nanotubes. Polymerization is the process in which
long-chain or network molecules are made from relatively small organic molecules.
D. Shi and P. He at UC have developed a unique plasma polymerization process to
coat CNT with different materials.
64
In a plasma environment, the surface atoms of
the CNT are activated. At these active centers, a polymer film forms on the CNT
surface. The polymerization conditions can be controlled so that the structure of the
coating contains the desired functional groups. High-resolution TEM pictures of
plasma-coated MWCNT are shown in Fig. 5.21. The coating thickness can be
controlled by the processing time and it is in the range of 20 nm. The reader can
refer to Chapter 8 of this book for more details on plasma coating of CNT.
Another approach for functionalization of CNT that we are currently investigat-
ing is based on deposition of a thin Pd coating. The goal of this research is to design
and fabricate a hydrogen nanosensor. When palladium absorbs H
2
, a new phase
(palladium hydride) is produced, causing the metal lattice to expand by about 3.5%.
As a result, the palladium grains within a thin Pd film swell.
69
The described effect
is used in a hydrogen sensor demonstrated by R. Penner and coworkers.
70
They
fabricated a miniature hydrogen sensor based on an array of Pd nanowires. Hydrogen
absorption lowers the Pd wire electrical resistance by changing the Pd structure.
When exposed to hydrogen, Pd wires expand because of H
2
absorption, and the
nanoscopic gaps between the wires close. This allows more current to pass through
the wires, which is proportional to the absorbed hydrogen.
We plan to coat columns of CNT, similar to those shown in Fig. 5.19, with
2030 nm Pd by electroless deposition. Sputtering is also considered as a route for
Pd metallization. The exposure of the Pd-coated CNT column to hydrogen will
FIGURE 5.20 Schematic drawing of arrangement for oriented grow of parallel CNT under
an electric field.
20 m
Negative
electrode
Positive
electrode
CNT
network
Catalyst
CVD reactor
change the electrical and mechanical properties of the coated CNT. The column is
expected to change the electrical resistance, the length, and also the direction of
orientation. In addition, the nanogaps between the individual CNT will shrink. This
process is reversible and the desorption of H
2
can be enhanced by slightly heating
the CNT sensor. Multiple cycles of hydrogen adsorption and desorption from the
Pd-coated CNT with predictable reproducibility are technically feasible.
5.8 CONCLUSIONS
The CVD technique is a powerful approach for synthesis of CNT with controlled
properties. It has been demonstrated that the manipulation of growth parameters
along with the right choice of catalyst can predetermine the deposition of either
SWCNT or MWCNT. The use of MgO-supported catalyst results in easy removal
of impurities by HCl acid.
Quadrupole mass spectroscopy can be employed for in situ monitoring of the CNT
growth environment in the CVD reactor. It was observed that this technique successfully
reveals changes in the concentration of the gas species during the pulsed hydrocarbon
delivery. The synthesis of CNT using pulse flow of the carbon precursor is a promising
approach. It is currently under investigation at UC, and new results will be published.
Advanced topics for CVD growth of CNT include (1) synthesis of long nanotubes
by using floating catalyst and vertical CVD reactor and (2) controlled alignment of
SWCNT selectively grown on flat substrate as arrays. These are the areas of interest to
be pursued in order to fabricate high-quality CNT for sensor and actuator applications.
FIGURE 5.21 HRTEM images of coated MWCNT showing horizontal view of one nanotube
with the catalyst particle on the top.
PROBLEMS
1. List the important benefits of the CVD approach as compared to other
preparation techniques for synthesis of CNT.
2. Perform a literature search on CVD growth of CNT and reveal any new
advances that have been made relative to the information provided in this
chapter.
3. List the catalysts, the substrates, and the carbon precursors used for
synthesis of CNT by CVD.
4. Describe the difference between the bulk synthesis and surface synthesis
in terms of yield and alignment of CVD-grown carbon nanotubes.
5. List the impurities of as-grown CNT and describe the purification proce-
dures for their removal.
6. Search the literature for new purification techniques that can decrease the
loss of CNT during the removal of the impurities.
7. List the characterization techniques used for determination of CNT diam-
eter. Comment on the limitations and advantages of each technique.
8. Describe the characterization techniques for studying the morphology and
the structure of carbon nanotubes.
ACKNOWLEDGMENTS
The work described in this chapter was sponsored in part by the UC Summer Student
Fellowship, the UC University Research Council, First Nano, Inc., and the Ohio
Aerospace Institute. We thank S. Subrahmin and R. Tiwari for the ESEM and TEM
images of CNT. The authors gratefully acknowledge C. Sloan, D. Adderton, and Y.
Tu of First Nano, Inc. for their cooperation with the nanotube synthesis. We appre-
ciate the high-resolution TEM images of SWCNT provided by L. Wang and J. Lian
from the University of Michigan and the Raman spectroscopy of CNT done by D.
Mast and R. Gilliland from UC and J. Busbee and J. Reber from AFRL/MLMT
Wright-Patterson AFB.
REFERENCES
1. Sullivan, J., Robertson, J., Zhou, O., Allan, T., and Coll, B., Amorphous and nano-
structured carbon, MRS Symposium Proceedings, 593, 2000.
2. Rao, A., Nanotubes and related materials, MRS Symposium Proceedings, 633, 2001.
3. Robertson, J., Friedmann, T., Geohegan, D., Luzzi, D., and Ruoff, R., Nanotubes,
fullerenes, nanostructured and disordered carbon, MRS Symposium Proceedings,
675, 2001.
. 4. Bernier, P., Roth, S., Carroll, D., and Kim, G.T., Nanotube-based devices, MRS
Symposium Proceedings, 772, 2003.
5. Dresselhaus, M. and Dai, H., Advances in carbon nanotubes, MRS Bulletin, 29, 4,
237, 2004.
6. Thess, A., Lee, R., Nikolaev, P., Dai, H., Petit, P., Robert, J., Xu, C., Lee, Y.H.,
Kim, S.G., Rinzler, A.G., Colbert, D.T., Scuseria, G.E., Tomnek, D., Fischer, J.E.,
and Smalley, R.E., Crystalline ropes of metallic carbon nanotubes, Science, 273,
483, 1996.
7. Bethune, D., Kiang, C., DeVries, M., Gorman, G., Savoy, R., and Beyers, R., Cobalt-
catalysed growth of carbon nanotubes with single-atomic-layer walls, Nature, 363,
605, 1993.
8. Journet, C., Maser, W.K., Bernier, P., Loiseau, A., Chapelle, M., Lefrant, S., Deniard,
P., Lee, R., and Fischer, J.E., Large-scale production of single walled carbon nano-
tubes by the electric-arc technique, Nature, 388, 756, 1997.
9. Cassel, A., Raymakers, J., Kong, J., and Dai, H., Large scale synthesis of single-
walled carbon nanotubes, J. Phys. Chem. B., 103, 6484, 1999.
10. Liu, J., Fan, S., and Dai, H., Recent advances in methods of forming carbon nanotubes,
MRS Bulletin, 29, 4, 224, 2004.
11. Smith, D., Thin-Film Deposition: Principles and Practice, 1st ed., McGraw-Hill, New
York, 1995.
12. Hitchman, M. and Jensen, K., Chemical Vapor Deposition: Principles and Applica-
tions, Academic Press, London, 1993.
13. Sherman, A., Chemical Vapor Deposition for Microelectronics: Principles, Technol-
ogy and Applications, Noyes Publications, NJ. 1998.
14. Kuang, M., Wang, Z., Bai, X., Guo, J., and Wang, E., Catalytically active nickel
{110} surfaces in growth of carbon tubular structures, Appl. Phys. Lett., 76, 10, 1255,
2000.
15. Lee, C. and Park, J., Growth model of bamboo-shaped carbon nanotubes by thermal
chemical vapor deposition, Appl. Phys. Lett., 77, 21, 3397, 2000.
16. Tang, C., Ding, X., Huang, X., Gan, Z., Qi, S., Liu, W., and Fan, S., Effective growth
of boron nitride nanotubes, Chem. Phys. Lett., 356, 254, 2002.
17. Li, W., Wen, J., and Ren, Z., Effect of temperature on growth and structure of carbon
nanotubes by chemical vapor deposition, Appl. Phys. A, 74, 397, 2002.
18. Li, W., Wen, J., Tu, Y., and Ren, Z., Effect of gas pressure on growth and structure
of carbon nanotubes by chemical vapor deposition, Appl. Phys. A, 73, 259, 2001.
19. Li, W., Wen, J., Sennett, M., and Ren, Z., Clean double-walled carbon nanotubes
synthesized by CVD, Chem. Phys. Lett., 368, 299, 2003.
20. Colomer, J., Stephan, C., Lefrant, S., Tendeloo, G., Willems, I., Konya, Z., Fonseca, A.,
Laurent, C., and Nagy, J., Large-scale synthesis of single-walled carbon nanotubes by
catalytic chemical vapor deposition (CCVD) method, Chem. Phys. Lett., 317, 83, 2000.
21. Geng, J., Singh, C., Shephard, D., Shaffer, M., Johnson, B., and Windle, A., Synthesis
of high purity single-walled carbon nanotubes in high yield, Chem. Commun., 2666,
2002.
22. Tu, Y., Lin, Y., and Ren, Z., Nanoelectrode arrays based on low site density aligned
carbon nanotubes, Nano Lett., 3, 1, 107, 2003.
23. Wen, J., Huang, Z., Wang, D., Chen, J., Yang, S., Ren, Z., Wang, J., Calvet, L., Klemic,
J., and Reed, M., Growth and characterization of aligned carbon nanotubes from pat-
terned nickel nanodots and uniform thin films, J. Mater. Res., 16, 11, 3246, 2001.
24. Bower, C., Zhou, O., Zhu, W., Werder, D., and Jin, S., Nucleation and growth of
carbon nanotubes by microwave plasma chemical vapor deposition, Appl. Phys. Lett.,
75, 12, 2767, 2000.
25. Fan, S., Chaplin, G., Franklin, N., Tombler, T., Cassell, A., and Dai, H., Self-oriented
regular arrays of carbon nanotubes and their field emission properties, Science, 283,
512, 1999.
26. Kong, J., Soh, H., Cassel, A., Quate, C., and Dai, H., Synthesis of individual single-
walled carbon nanotubes on patterned silicon wafers, Nature, 395, 878, 1998.
27. Lee, C., Kim, D., Lee, T., Choi, Y.C., Park, Y., Kim, W., Lee, Y., Choi, W., Lee, N.,
Kim, J., Choi, Y.G., and Kim, J., Synthesis of uniform distributed carbon nanotubes
on a large area of Si substrates by thermal chemical vapor deposition, Appl. Phys.
Lett., 75, 12, 1721, 1999.
28. Chakrapani, N., Curran, S., Wei, B., Ajayan, P., Carrillo A., and Kane, R., Spectral
fingerprinting of structural defects in plasma-treated carbon nanotubes, J. Mater.,
Res., 18, 10, 2515, 2003.
29. Zhu, H., Xu, C., Wu, D., Wei, B., Vajtai, R., and Ajayan, P., Direct synthesis of long
single-walled carbon nanotube strands, Science, 296, 884, 2002.
30. Ci, L., Li, Y., Wei, B., Liang, J., Xu, C., and Wu, D., Preparation of carbon nanofibers
by the floating catalyst method, Carbon, 38, 1933, 2000.
31. Singh, C., Quested, T., Boothroyd, C., Thomas, P., Kinloch, I., Kandil, A., and Windle,
A., Synthesis and characterization of carbon nanofingers produced by the floating
catalyst method, J. Phys. Chem. B, 106, 10915, 2002.
32. First Nano, Inc., EasyTube Users Manual, Preliminary Version 10/16/02, Santa
Barbara, CA, 2002.
33. Shanov, V., Ramamurti, R., and Singh, R.N., Effect of Pulsed Methane Flow on
nanostructured diamond films synthesized by microwave plasma CVD, Transactions,
154, 283, 2003.
34. Hafner, J., Bronikowski, M., Azamian, B., Nikolaev, P., Rinzler, A., Colbert, D.,
Smith, K., and Smalley, R., Catalytic growth of single-wall carbon nanotubes from
metal particles, Chem. Phys. Lett., 296, 195, 1988.
35. Fan, Q., Ali, N., Kousar, Y., Ahmed, W., and Gracio, J., Novel time-modulated chemical
vapor deposition process for growing diamond films, J. Mater. Res., 17, 17, 2002.
36. Ali, N., Neto, V.F., and Gracio, J., Promoting secondary nucleation using methane
modulations during diamond chemical vapor deposition to produce smoother, harder,
and better quality films, J. Mater. Res., 18, 2, 2003.
37. Silva, F., Gicquel, A., Chiron, A., and Achard, J., Low roughness diamond films
produced at temperatures less than 600qC, Diamond Relat. Mater., 9, 12, 1965, 2000.
38. Ramamurti, R., Shanov, V., and Singh, R.N., Synthesis of Nanocrystalline Diamond
Films by Microwave Plasma CVD, in Proc. Ceramic transactions, Bansal, NP., and
Singh, JP., Eds., The American Ceramic Society, OH, 135, 39, 2003.
39. Ramamurti, R., Shanov, V., and Singh, R. N., Processing of Nanocrystalline Diamond
Films by Microwave Plasma CVD, in Ceramic Engineering and Science Proceedings,
Kriven, W., and Lin, H.T., Eds., The American Ceramic Society, OH, 24(3), 15, 2003.
40. Jain, S., Kang, P., Yun,Y., He, P., Pammi, S., Miskin, A., Narsimhadevara, S., Hurd,
D., Schulz, M., Chase, J., Subramaniam, S., Shanov, V., Boerio, F.J., Shi, D., Gilliland,
R., Mast, D., and Sloan, C., Building Smart Materials Based on Carbon Nanotubes,
in Proc., SPIE, Varadan, V.K., Eds., Society of Photo-Optical Instrumentation Engi-
neers, Bellingham, WA, 5389, 167, 2004.
41. Haddon, R., Sippel, J., Rinzler, A., and Papadimitrakopoulos, F., Purification and
separation of carbon nanotubes, MRS Bulletin, 29, 4, 252, 2004.
42. Huang, S. and Dai, L., Plasma etching for purification and controlled opening of
aligned carbon nanotubes, J. Phys. Chem. B, 106, 3543, 2002.
43. Hou, P., Bai, S., Yang, Q., Liu, C., and Cheng, H., Multi-step purification of carbon
nanotubes, Carbon, 40, 81, 2001.
44. Zheng, B., Li, Y., and Liu, J., CVD synthesis and purification of single-walled carbon
nanotubes on aerogel-supported catalyst, Appl. Phys. A, 74, 345, 2002.
45. Hernadi, K., Fonseca, A., Nagy, J., Bernaerts, D., Riga, J., and Lucas, A., Catalytic
synthesis and purification of carbon nanotubes, Synth. Met., 77, 31, 1996.
46. Dillon, A., Gennett, T., Jones, K., Alleman, J., Parilla, P., and Heben, M., A simple
and complete purification of single-walled carbon nanotube materials, Advanced
Mater., 11, 16, 1354, 1999.
47. Sloan, J., Luzzi, D., Kirkland, A., Hutchison, J., and Green, M., Imaging and char-
acterization of molecules and one-dimensional crystals formed within carbon nano-
tubes, MRS Bulletin, 29, 4, 265, 2004.
48. Singh, C., Shaffer, M., Kozoil, K., Kinloch, I., and Wndle, A., Towards the production
of large-scale aligned carbon nanotubes, Chem. Phys. Lett., 372, 860, 2003.
49. Loehman, R., Characterization of Ceramics, Butterworth-Heinemann, Boston, 993,
284.
50. Jiao, J., Einarsson, E., Tuggle, D., Love, L., Prado J., and Coia, G., High-yield
synthesis of carbon coils on tungsten substrates and their behavior in the presence of
an electric field, J. Mater. Res., 18, 11, 2580, 2003.
51. Lewis, I. and Edwards, H., Handbook of Raman Spectroscopy: From the Research
Laboratory to the Process Line, Marcel Dekker, New York, 2001.
52. Szymanski, H., Raman Spectroscopy: Theory and Practice, Plenum Press, New York,
1967
53. McCreery, R., Raman Spectroscopy for Chemical Analysis, John Wiley & Sons, New
York, 2000.
54. Bachmann, K. and Wiechert, D.U., Characterization and properties of artificially
grown diamond, in Clausing, R., Horton, L., Angus, J., and Koidl, P., Eds., Diamond
and Diamond-like Films and Coatings, Plenum Press, New York, 677, 1991.
55. Eklund, C., Holden, M., and Jishi, A., Vibrational modes of carbon nanotubes:
Spectroscopy and theory, Carbon, 33, 7, 959, 1995.
56. Rao, A., Richter, E., Bandow, S., Chase, B., Eklund, P., Williams, K., Fang, S.,
Subbaswamy, K., Menon, M., Thess, A., Smalley, R., Dresselhaus, G., and Dresselhaus
M.S., Diameter-selective Raman scattering from vibrational modes in carbon nano-
tubes, Science, 275, 5297, 187, 1997.
57. Tan, S., Zhang, P., Yue, K., Huang, F., Shi, Z., Zhou, X., and Gu, Z., Comparative
study of carbon nanotubes Raman prepared by D.C. arc discharge and catalytic
methods, J. Raman Spectrosc., 28, 369, 1997.
58. Bandow, S., Asaka, S., Sato, Y., Grigorian, L., Richter, E., and Ecklund, P., Effect of
the growth temperature on the diameter distribution and chirality of single-wall carbon
nanotubes, Phys. Rev. Lett., 80, 17, 3779, 1998.
59. Saito, R., Takeya, T., Kimura, T., Dresselhaus, G., and Dresselhaus, M.S., Finite-
size effect on the Raman spectra of carbon nanotubes, Phys. Rev. B, 59, 31, 2388,
1999.
60. Benoit, J., Buisson, J., Chauvet, O., Godon, C., and Lefrant, S., Low-frequency Raman
studies of multiwalled carbon nanotubes: Experiments and theory, Phys. Rev. B, 66,
073417/1, 2002.
61. Alvarez, L., Righi, A., Rols, S., Anglaret, E., and Sauvajol, J.L., Excitation energy
dependence of the Raman spectrum of single-walled carbon nanotubes, Chem. Phys.
Lett., 320, 441, 2000.
62. Lin, Y., Lu, F., Tu, Y., and Ren, Z., Glucose biosensors based on carbon nanotube
nanoelectrode ensembles, Nano Lett., 4, 2, 191, 2004.
63. Bronikowski, M. and Hunt, B., Block copolymers as templates for arrays of carbon
nanotubes, NASA Tech Briefs, 5657, 2003.
64. Shi, D., Lian, J., He, P., Wang, L., van Ooij, W., Schultz, M., and Mast, D., Plasma
deposition of ultrathin polymer films on carbon nanotubes, Appl. Phys. Lett., 81, 27,
5216, 2002.
65. Zhang, Y., Chang, A., Cao, J., Qian, Q., Woong, W., Li, Y., Yiming, Morris, N.,
Yenilmez, E., Kong, J., and Dai, H., Electric-field-directed growth of aligned single-
walled carbon nanotubes, Appl. Phys. Lett., 79, 19, 3155, 2001.
66. Cassell, A., Franklin, N., Tombler, T., Chan, E., Han, J., and Dai, H., Directed growth
of free-standing single-walled carbon nanotubes, J. Am. Chem. Soc., 121, 7975, 1999.
67. Huang, S., Cai, X., and Liu, J., Growth of millimeter-long and horizontally aligned
single-walled carbon nanotubes on flat substrates, J. Am. Chem. Soc., 125, 19, 5636,
2003.
68. Khare, B. and Meyyappan, M., Low-temperature plasma functionalization of carbon
nanotubes, NASA Tech Briefs, 45, 2004.
69. Lewis, F., The Palladium Hydrogen System, Academic Press, New York, 1967.
70. Favier, F., Walter, E., Zach, M., Benter, T., and Penner, R., Hydrogen sensor and
switches from electrodeposited palladium mesowire arrays, Science, 293, 2227, 2001.
6
Self-Assembled Au
Nanodots in a ZnO
Matrix: A Novel Way
to Enhance Electrical
and Optical
Characteristics
of ZnO Films
Ashutosh Tiwari and Jagdish Narayan
CONTENTS
6.1 Introduction
6.1.1 Nanostructured Materials
6.1.2 Transparent and Conducting ZnO
6.1.3 Au Nanodots in a ZnO Matrix
6.2 Experimental Procedure
6.2.1 Pulsed Laser Deposition
6.2.2 Experimental Setup
6.2.3 Target Arrangement
6.2.4 Growth Parameters
6.2.5 Characterization
6.3.1 Structural
6.3.2 Optical
6.3.3 Electrical
6.4 Conclusion
Problems
Acknowledgments
References
6.1 INTRODUCTION
6.1.1 NANOSTRUCTURED MATERIALS
During the last few years nanostructured materials and devices have attracted a great
deal of scientic and technological attention.
13
The properties of nanostructures
are strongly inuenced by quantum effects. They exhibit novel optical, magnetic,
and electronic properties, and present the perspective of designing and tuning mate-
rials properties with exceptional exibility and control. Nanostructured materials are
of enormous interest, from the point of view of discovering new physical phenomena
as well as their exploitation possibilities in novel devices. As the size and dimen-
sionality of a material decrease, its electronic structure and properties change tre-
mendously due to the reduction in the density of states and the effective length scale
of electronic motion. Now the energy levels of material are determined mostly by
the materials surface. This results in a transition from the bulk band structure to
individual localized energy levels and hence the onset of quantum connement
effects. Because of the size-induced change in the electronic structure, optical prop-
erties of nanomaterials also change signicantly. In the case of nanoparticles of
noble metals, the surface plasmon resonance becomes very important and results in
strong absorption of light in characteristic frequency range. Surface plasmon reso-
nance occurs because of the coherent oscillations of the conduction band electrons
induced by interaction with an electromagnetic eld. The electric eld of the incident
light beam induces a polarization of the electrons with respect to the much heavier
ionic core of the nanoparticles. A net charge difference is felt at the nanoparticle
surface, which in turn acts as a restoring force. This creates a dipolar oscillation of
all the electrons with the same phase. When the frequency of the electromagnetic
eld becomes resonant with the coherent electron motion, a strong absorption of
incident radiation occurs. In the case of noble metals, particularly gold (Au) nan-
odots, a strong surface plasmon resonance occurs in the visible part of the electro-
magnetic radiation. Actual frequency and width of the surface plasmon resonance
absorption depend on the size and shape of nanoparticles and the dielectric constant
of metal and surrounding matrix. In this chapter we will show that how the phe-
nomenon of surface plasmon resonance observed in gold nanodots can be utilized
to enhance the optical and electrical characteristics of zinc oxide (ZnO).
6.1.2 TRANSPARENT AND CONDUCTING ZnO
Zinc oxide is one of the most promising materials in the fast-emerging optoelec-
tronics industry.
46
It has a band gap of 3.27 eV at 300 K, which is very close to
the band gap of 3.44 eV for GaN.
7
In addition, it has higher exciton binding energy
of 60 meV, compared to 25 meV for GaN. High exciton binding energy results in
less trapping of carriers, and therefore higher luminescent efciencies are expected.
Zinc oxide lms are also being considered for ohmic contacts to GaN due to their
high electrical conductivity and matching band gap with GaN. Therefore, one of the
most challenging issues in ZnO-related research is to increase the electrical conduc-
tivity of ZnO without compromising its optical characteristics.
8,11
It has been realized
that any effort to increase the conductivity of ZnO by doping or by creating oxygen
deciencies always results in the deterioration of the optical efciency of the lm.
8,9,12
These dopants and vacancies create defect levels, and hence a signicant reduction
in the band-edge PL peak is observed along with a very broad green band in PL
spectrum.
13,14
6.1.3 AU NANODOTS IN A ZnO MATRIX
In this chapter we show that both the electrical and optical characteristics of ZnO
can be improved by carefully embedding nanosized epitaxial Au particles in it. We
have shown that ZnO-Au epitaxial composite lms exhibit higher electrical con-
ductivity and negligible green band emission in photoluminescence spectra. The
choice of Au in this system was inspired by the following facts: (1) Au has negligible
solubility, and therefore does not affect the chemical composition of the host ZnO
matrix, and it grows epitaxially within the ZnO matrix; (2) Au has virtually no
afnity for oxygen and hence retains its high conductivity, which results in overall
enhancement in electrical conductivity of the composite lm; and (3) the most
important reason for this selection was that for Au, plasmon resonance occurs at
around 520 nm,
15
which lies near the green band emission regime of ZnO. In the
case of an assembly of nanosized metal particles with a size distribution, the
plasmon resonance occurs over a range of frequencies. The exact value of this
resonance frequency range is determined by the dielectric properties of the matrix
and size distribution of nanoparticles. By carefully tailoring the size distribution
of Au nanodots, it is possible to extend the plasmon resonance spectrum to the
whole green band emission regime. Under this condition the green light radiation
originating from the midgap impurity states of ZnO can be suppressed by Au
nanodots, leaving behind pure excitonic emission of ZnO.
6.2 EXPERIMENTAL PROCEDURE
6.2.1 PULSED LASER DEPOSITION
We used pulsed laser deposition (PLD) technique to embed Au nanodots in a ZnO
matrix. PLD is an efcient method to produce thin lms and structures by utilizing
the concept of laser-solid interaction. The principle of PLD is quite simple. A highly
intense UV laser beam is focused on a target where the high energy density during
the laser pulse (about 1 GW within 25 ns) ablates the material. The ablated material
forms a plasma that is deposited on a substrate opposite the target. PLD technique
can be used for almost any material, particularly for compounds that are difcult or
impossible to produce in thin-lm form by other techniques.
6.2.2 EXPERIMENTAL SETUP
The experimental setup used in our method consists of a pulsed excimer laser,
multitarget stainless steel chamber, and a turbomolecular pump (Fig. 6.1). A vacuum
is maintained in the chamber by means of a port to the turbomolecular pump. Targets
are positioned parallel to a substrate that is placed on a heater plate. The distance
between the target and the substrate is kept at about 5 cm. We used a pulsed KrF
excimer laser (wavelength 248 nm) to ablate the targets.
6.2.3 TARGET ARRANGEMENT
In order to grow nanocrystalline Au dots embedded in a ZnO matrix, high-purity
Au and ZnO targets were alternately ablated by arranging the targets in a special
geometrical conguration (Fig. 6.2). In this conguration, a small strip of Au was
glued to a circular pallet of ZnO and a pulsed laser beam was made to hit this
assembly at 45 from the normal. The whole assembly was rotated with a motor,
which results in alternate ablation of ZnO and Au targets.
6.2.4 GROWTH PARAMETERS
Key parameters for controlling the size of Au nanodots and distribution were (1)
relative time of ablation of Au and ZnO (i.e., the relative area of ablation of these
targets) and (2) energy density and repetition rate of the laser beam. These deposition
parameters were carefully controlled to obtain the desired size and distribution of
Au particles. Films with optimum electrical and optical performance were prepared
FIGURE 6.1 Schematic diagram of pulsed laser deposition system.
Motor
View window
View window
Target
Substrate
Thermocouple
Substrate
heater
Heater wires
Quartz lens
Mirror
Aperture
Quartz window
Gas pumpout
port
Excimer laser
by using the following combination of deposition parameters: relative area of abla-
tion of Au and ZnO = 1:4, laser energy density = 1.5 J/cm
2
, and repetition rate =
15 Hz. All the lms were deposited on (0001) plane of sapphire at 650
o
C under the
oxygen ambient of 1 10
4
torr. Under these conditions the deposition rate of ZnO-
Au composite lm was about 10 nm/min.
6.2.5 CHARACTERIZATION
Chemical composition, size distribution, and microstructure of nanodots and sur-
rounding matrix were determined by using cross-sectional scanning transmission
electron microscopy with atomic number (Z) sensitive contrast (STEM-Z). Photo-
luminescence measurements were performed using a Hitachi F-2500 spectropho-
tometer. Electrical resistivity of these lms was measured by the four-point probe
method in the temperature range of 12300 K, using a Kiethely Current source
(Model 220) and a nanovoltmeter (Model-182).
6.3.1 STRUCTURAL
Figure 6.3 shows the x-ray diffraction pattern of ZnO-Au composite lm grown on
sapphire substrate. In the 2 range (2080), we observed only the peaks corre-
sponding to (0002) and (0004) planes of ZnO, (0006) plane of sapphire, and (111)
plane of Au.
Shown in Fig. 6.4(a) is a typical cross-sectional transmission electron micros-
copy image of one of the composite lms. Uniformly distributed nanodots can clearly
be seen. These nanodots are almost spherical in shape with a radius of 26 nm.
They are distributed in the matrix with an average spacing of 412 nm. High-
resolution TEM image of one typical Au nanodot is shown in Fig. 6.4(b). The Au
nanodots grow epitaxially within the ZnO matrix via domain matching epitaxy,
16
where integral multiples of lattice planes match across the Au-ZnO interface.
FIGURE 6.2 Geometrical arrangement of ZnO and Au targets for depositing ZnO-Au com-
posite lms.
Laser
beam
Au
ZnO
6.3.2 OPTICAL
The transmission and absorption spectra of ZnO-Au composite lm is shown in Fig.
6.5. Apart from the band-edge absorption of ZnO at ~375 5 nm, we also observed
a broad absorption peak corresponding to plasmon resonance spectrum of Au nan-
odots. The shape and the position of this broad absorption peak was found to be a
function of particle size and its distribution. For the distribution of Au particles
shown in Fig. 6.4(a), plasmon absorption peak lies in the wavelength range of
450600 nm with a maxima at 522 nm.
FIGURE 6.3 X-ray diffraction pattern of ZnO-Au composite lm.
FIGURE 6.4 Cross-sectional TEM image of ZnO-Au composite lm showing uniformly
distributed Au nanodots (a), and high-resolution TEM image of one typical Au nanodot (b).
4000
2000
20 40 60 80
I
n
t
e
n
s
i
t
y

(
a
r
b
.

u
n
i
t
s
)
2 (deg.)
Z
n
O
(
0
0
0
2
)
Z
n
O
(
0
0
0
4
)
A
u
(
1
1
1
)
-
A
l
2
O
3
a b
20 nm
2 nm
Because of this absorption peak, the intensity of any green band emission
originating from the defect levels of ZnO is expected to be strongly damped. Shown
in Fig. 6.6 is the room temperature photoluminescence of ZnO-Au composite lms.
This spectrum shows only one sharp peak at ~385 5 nm. For a comparison, we
have also shown the photoluminescence spectrum of ZnO lms prepared under the
FIGURE 6.5 Transmission and absorption spectra of ZnO-Au composite lm.
FIGURE 6.6 Photoluminescence spectra of ZnO-Au composite lm (a), and photolumines-
cence spectra of ZnO lm prepared under the exact same conditions as the composite lm (b).
2.0
1.6
1.2
0.8
0.4
A
b
s
o
r
b
a
n
c
e
800 600
450 600 750
400
30
20
10
0
T
r
a
n
s
m
i
t
t
a
n
c
e

(
%
)
(nm)
(nm)
800
400
800
400
0
0
I
n
t
e
n
s
i
t
y

(
a
r
b
.

u
n
i
t
s
)
400 500 600 700
(a)
(b)
(nm)
exact same conditions. Presence of a broad green band emission peak for bare ZnO
lm and its complete disappearance in the ZnO-Au composite lm are clearly
demonstrated.
6.3.3 ELECTRICAL
Figure 6.7(a) shows the variation of electrical resistivity of the ZnO-Au composite
with temperature. At room temperature the value of electrical resistivity is 3.4 m-
cm, and it increases with the decrease in the temperature. The inset in this gure
shows a plot of ln vs. 1/T. This graph exhibits a sharp change in the slope of ln
vs. 1/T at around 150 K. This suggests that the system exhibits two different values
of excitation energies at low and high temperatures. Slope of ln vs. 1/T at low
temperature corresponds to an excitation energy of 0.2 meV, while that at higher
temperature (T > 150 K) corresponds to the excitation energy of 1 meV. It is
interesting to note that these energies are much smaller than the energies usually
FIGURE 6.7 Electrical resistivity vs. temperature for ZnO-Au composite lm. The inset
shows the plot of ln vs. 1/T (a). Electrical resistivity versus temperature for pure ZnO lm (b).
5.5
5.2
5.4
5.6
5.0
4.5
4.0
3.5
800
600
400
200
50 100 150 200 250 300
T (K)

(
m
-
c
m
)

(
m
-
c
m
)
I
n

(
)
0.08 0.16 0.24
1/T (K
1
)
(a)
(b)
observed in ZnO systems, even with very shallow donor levels. For comparison we
also measured the electrical resistivity of bare ZnO lm prepared under the exact
same conditions (Fig. 6.7(b)). This lm exhibits activation energy of 40 meV, which
is in good agreement with previous reports. The extremely low values of activation
energies for composite lms suggest that the simple activated conduction mechanism
cannot account for the observed behavior of electrical resistivity. However, the fact
that at lower temperatures the estimated activation energy is smaller suggests that
in this temperature regime electrostatic charging of nanodots provides another chan-
nel of electrical conduction.
6.4 CONCLUSION
In this chapter we have presented a novel approach to embed nanosized Au particles
in epitaxial ZnO lms via a pulsed laser deposition technique. Au-embedded epi-
taxial ZnO lms were deposited on (0001) sapphire substrates. The crystalline
quality of both the ZnO matrix and Au nanoparticles were investigated by x-ray
diffraction and transmission electron microscopy. Composite lms were character-
ized by photoluminescence, optical absorption, and low temperature electrical resis-
tivity measurements. It has been shown that Au particles can suppress the green
band emission of ZnO via the process of energy transfer from defect levels to the
surface plasmon of Au. These ZnO-Au lms exhibit high electrical conductivity
with room temperature resistivity of 3.4 0.2 m-cm and a sharp excitonic peak
at 3.22 0.05 eV without any green band emission. Growth of highly conducting
ZnO-Au composite lms with improved optical characteristics is an important step
in the fast-emerging optoelectronics industry.
PROBLEMS
1. Describe and discuss surface plasmon resonance.
2. What are the values of band gap and exciton binding energies for ZnO?
Compare these numbers with corresponding values for GaN.
ACKNOWLEDGMENTS
The authors gratefully acknowledge nancial support from the National Science
Foundation the U.S. Army Research Laboratory, and the U.S. Army Research Ofce.
We also thank Dr. C. Jin and Mr. Amit Chugh for useful discussion.
REFERENCES
1. Narayan, J. and Tiwari, A. Methods of Forming Three-Dimensional Nanodot Arrays
in a Matrix. U.S. Patent Application # 60/430, 210 (December 2, 2002).
2. Narayan, J. and Tiwari, A. Novel Methods of Forming Self-Assembled Nanostruc-
tured Materials: Ni Nanodots in Al
2
O
3
and TiN Matrices. Journal of Nanoscience
and Nanotechnology, 4, 726, 2004.
3. Tiwari, A., Jin, C., Chugh, A., and Narayan, J. Role of Self-Assembled Gold
Nanodots in Improving the Electrical and Optical Characteristics of Zinc Oxide
Films. Journal of Nanoscience and Nanotechnology, 3, 368 (2003).
4. Ohno, H. Making Nonmagnetic Semiconductors Ferromagnetic. Science, 281, 951
(1998).
5. Yu, P., Tang, Z.K., Wong, G.K.L., Kawasaki, M., Ohtomo, A., Koinuma, H., and
Segawa, Y. Ultraviolet Spontaneous and Stimulated Emissions from Zno Microc-
rystallite Thin Films at Room Temperature. Solid State Commun., 103, 459 (1997).
6. Tiwari, A., Jin, C., Kvit, A., Kumar, D., Muth, J.F., and Narayan, J. Structural,
Optical and Magnetic Properties of Diluted Magnetic Semiconducting Znl-Xmnxo
Films. Solid State Commun., 121, 371 (2002).
7. Viswanath, A.K., in Handbook of Advanced Electronic and Photonic Materials and
Devices, edited by Nalwa, H.S., Academic Press, San Diego, CA (2001), Vol. 1, p. 109.
8. Hong, R.J., Jiang, X. Heide, G., Szyszka, B., Sittinger, V., and Werner, W. Growth
Behaviours and Properties of the ZnO: Al Films Prepared by Reactive Mid-Frequency
Magnetron Sputtering. Journal of Crystal Growth, 249, 461 (2003).
9. Jin, B.J., Woo, H.S., Im, S., Bae, S. H., and Lee, S.Y. Relationship Between Pho-
toluminescence and Electrical Properties of Zno Thin Films Grown by Pulsed Laser
Deposition. Applied Surface Science, 169, 521 (2001).
10. Exarhos, G.J., Rose, A., and Windisch, C.F. Spectroscopic Characterization of
Processing-Induced Property Changes in Doped ZnO Films. Thin Solid Films, 308,
56 (1997).
11. Bolzlee, B.J. and Exarhos, G.J. Preparation and Characterization of Gold and Ruthe-
nium Colloids in Thin Zinc Oxide Films. Thin Solid Films, 377, 1 (2000).
12. Reynolds, D.C., Look, D.C., and Jogai, B. Find Structure on the Green Band in
Zno. Journal of Applied Physics, 89, 6189 (2001).
13. Reynolds, D.C. Look, D.C., Jogai, B., and Morcoc, H. Similarities in the bandedge
and Deep-Centre Photluminescence Mechanisms of ZnO and GaN. Solid State
Commun., 101, 643 (1997).
14. Tiwari, A., Park, M., Jin, C., Wang, H., Kumar, D., and Narayan, J. Epitaxial Growth
of Zno Films on Si(111). Journal of Materials Research, 17, 2480 (2002).
15. Averitt, R.D., Westcott, S.L., and Halas, N.J. Linear Optical Properties of Gold
Nanoshells. Journal of Optical Society of America B-Optical Physics, 16, 1824,
1999.
16. Narayan, J. and Larson, B.C. Domain Epitaxy: A Unied Paradigm for Thin Film
Growth. Journal of Applied Physics, 93, 278 (2003); J. Narayan, U.S. Patent 5, 406,
123 (4/11/1995).
17. Han, J., Senos, A.M.R., Mantas, P.Q., and Cao, W. Dielectric Relaxation of
Shallow Donor in Polycrystalline Mn-Doped ZnO. Journal of Applied Physics, 93,
4097, 2003.
18. Alivisatos, A.P. Semiconductor Clusters, Nanocrystals, and Quantum Dots. Science,
271, 933, 1996.
7
Synthesis of Boron
Nitride Nanotubes
Using a Ball-Milling
and Annealing Method
Ying Chen and Jim S. Williams
CONTENTS
7.1 Introduction
7.1.1 Boron Nitride Nanotubes
7.1.2 High-Energy Ball Milling (HEBM)
7.2 Synthesis of BN Nanotubes from Elemental B
7.2.1 Milling with a Rotating Steel Mill
7.2.2 Milling with a Vibrating Ceramic Mill
7.3 Synthesis of BN Nanotubes from BN Compounds
7.4 Formation Mechanism Discussion
7.4.1 Separate Nucleation and Growth
7.4.2 Nucleation Structures
7.4.3 The Growth Process
7.5 Summary
Problems
Acknowledgments
References
7.1 INTRODUCTION
Boron nitride (BN) nanotubes have a structure similar to carbon (C) nanotubes but
exhibit many interesting electronic and chemical properties. This chapter summarizes
research in the synthesis of BN nanotubes using a new ball-milling and annealing
method, which consists of high-energy ball milling and then isothermal annealing
processes. The high-energy ball milling creates a precursor for BN nanotube forma-
tion during subsequent annealing. This review of BN nanotube synthesis under
different milling and annealing conditions outlines new mechanisms of nanotube
nucleation and growth involved in the milling-annealing method.
7.1.1 BORON NITRIDE NANOTUBES
Boron nitride (BN) and carbon (C) share many common structures such as graphite
and hexagonal BN, and diamond and cubic BN crystalline structures. After the
discovery of C nanotubes in 1991,
1
BN nanotubes were rst predicted theoretically
2
and then successfully synthesized using the same arc-discharge method by Chopra
et al. in 1995.
3
Similar to C nanotubes, BN nanotubes can be simply regarded as
a cylinder rolled up from a graphitic BN sheet with a diameter less than 100 nm
and a length in the micrometer range.
4
However, there are two main differences
between structures of BN and C nanotubes. First, in the nanotube tips a single-wall
C nanotube is capped by a half-fullerene molecule (C
60
) consisting of six-membered
hexagons and ve-membered pentagons, and thus C nanotubes have cone-shape
tips (Fig. 7.1(a)).
5,6
However as shown in Fig. 7.1(b), most BN nanotubes have at
tips, which are formed with four-membered squares instead of ve-membered
pentagons.
79
Theoretical investigation reveals that ve-membered pentagons in BN
would require energetically unfavorable B B or N N bonds that destabilize the
structure.
10,11
A second difference is in chirality. Chirality, the dominant BN nano-
tube structures, are so-called zigzag tubes.
12,13
Armchair and chiral tubes are less
likely, possibly due to the special tip congurations.
10
On the other hand, C nanotubes
with zigzag, armchair, and chiral structures are all particularly abundant.
14
In addition
to typical parallel-walled cylindrical tubes, bamboo- and cone-like BN nanotubes
can also be produced.
15,16
Similar to C nanotubes, BN nanotubes can be lled with
metals (Fe, Ni, W, . . .)
3,8,17,18
and buckyballs.
19
Due to the similar tubular structure and ultrastrong sp
2
B N bonds, BN nano-
tubes have similarly superstrong mechanical properties as C nanotubes, with a
Youngs modulus of 1.22 0.24 TPa.
20
Theoretical studies suggest that BN nano-
tubes are stable insulating materials with a uniform electronic band gap of about
5.5 eV, which does not depend on size and chirality of nanotubes as in the C
FIGURE 7.1 TEM images of (a) a multiwalled C nanotube with a cone-shape tip and (b) a
BN nanotube with a at tip, indicated by arrow.
nanotube case.
21
BN nanotubes have a much better resistance to oxidation at high
temperatures than C nanotubes.
22
C nanotubes readily oxidize in air at a temperature
of 400C, starting from the tips and defects on the outer layers. They burn com-
pletely at 700C when sufcient oxygen is supplied.
23,24
We studied oxidation
behavior of C and BN nanotubes using thermal gravimetric analysis. The results
are illustrated in Fig. 7.2, which shows weight changes of nanotube samples as a
function of heating temperature in an air ow.
22
In the case of BN nanotubes (Fig.
7.2(a)), a sharp weight increase from 800 to 1100C, which is caused by the
FIGURE 7.2 Thermal gravimetric curves of sample weight changes as a function of temper-
ature for (a) BN nanotubes and (b) C nanotubes. The samples were heated at 20C min
1
in
an air ow of 10 ml min
1
.
22
0 200 400 600 800 1000 1200
29
30
31
32
33
34
(a)
S
a
m
p
l
e

w
e
i
g
h
t

(
m
g
)
Temperature (C)
BN + O
2
B
2
O
3
+ N
2
Fe + O
2
Fe
2
O
3
0 200 400 600 800 1000 1200
0.0
0.5
1.0
1.5
2.0
2.5
3.0
(b)
S
a
m
p
l
e

w
e
i
g
h
t

(
m
g
)
Temperature (C)
C + O
2
CO
2
oxidation of BN with the formation of solid boron oxides, is observed. In the case
of C nanotubes (Fig. 7.2(b)), a rapid weight loss due to oxidation with releasing
carbon gaseous oxides indicates that the oxidation starts from 400C and is complete
at 700C, consistent with literature reports.
23,24
Transmission electron microscopy
(TEM) investigation conrmed that BN nanotubes with different nanostructures
(cylinders and bamboos) and of a large range of sizes (2200 nm) are stable in air
at 700C, at which temperature C nanotubes completely burn off. Thin BN nano-
tubes with perfect nanocrystalline structures can withstand temperatures of up to
900C.
22
Thus, the superstrong mechanical properties and high resistance to oxi-
dation make BN nanotubes ideal for composite materials applications and the stable
electronic structure is useful in device applications.
The rst successful synthesis of pure BN nanotubes was made by Zettls group
in 1995 using an arc-discharge process similar to that used for carbon nanotube
production, where the tungsten anode contained an inserted BN rod rather than
graphite.
3
Loiseau et al.
7
and Terrones et al.
8
conrmed the formation of BN
nanotubes using the arc-discharge process with different metal catalysts. A high
yield of BN nanotubes was obtained by using carbon nanotubes as templates, where
C was substituted by boron and nitrogen atoms through a chemical reaction between
boron oxide vapor and nitrogen gas at elevated temperatures.
25
A low-temperature
oxidation is needed to remove residual C inside BN nanotubes.
26
Lee et al.
13
reported
the production of gram quantities of single-walled nanotubes, and pure double-
walled BN nanotubes were produced by Cunning et al.
27
However, all of the pre-
ceding deposition methods present difculties in scaling up for an industrial process.
Alternatively it was rst demonstrated in 1999 that a milling-annealing process, in
which boron powder was rst milled in ammonia gas and followed by annealing
in nitrogen atmosphere, had the potential to produce large quantities of BN nano-
tubes.
28
For example, up to 1 kg amounts of BN nanotubes are now possible using
a laboratory-scale milling device, thus opening up studies of new properties and
applications of BN nanotubes.
29
New formation mechanisms involved with such
high-yield BN nanotube formation have been discovered, as outlined in the following
sections.
7.1.2 HIGH-ENERGY BALL MILLING (HEBM)
Conventional ball milling is a traditional powder-processing technique that is widely
used in mineral, pharmaceutical, and ceramic industries. Small grinding machines
are popular tools in scientic laboratories for preparing ne powders for various
analyses and applications. However, high-energy ball milling (HEBM), rst
reported by Benjamin et al.
30
in the 1960s, is fundamentally different from the
conventional ball-milling technique. For example, a much more powerful attritor ball
mill had to be developed to synthesize new oxide-dispersion-strengthened alloys since
conventional ball mills could not provide sufcient grinding energy.
30
Intensive
research in synthesis of new metastable materials by HEBM was stimulated by the
pioneering work of amorphization of the Ni-Nb system conducted by Kock et al. in
1983.
31
Since then, a wide spectrum of metastable materials has been produced,
including nanocrystalline,
32
nanocomposite,
33
nanoporous phases,
34
supersaturated
solid solutions,
35
and amorphous alloys.
36
These new phase transformations induced
by HEBM are generally referred to as mechanical alloying (MA). At the same time,
it was found that HEBM can activate at room temperature chemical reactions, which
are normally only possible at high temperatures.
37
This latter process is called reactive
milling or mechanochemistry. Reactive ball milling has produced a large range of
nanosized oxides,
38
nitrides,
39
hydrides,
40
and carbide
41
particles. The major differ-
ences between conventional ball milling and high-energy ball milling are listed in
Table 7.1. The impact energy of HEBM is typically 1000 times higher than conven-
tional ball milling energy. The particle fracturing and size reduction in the conven-
tional ball milling is only the rst stage of HEBM. A longer milling time is thus
required for HEBM. In addition to the milling intensity, milling atmosphere and
temperature controls of HEBM are crucial to create a desired structural change or
chemical reaction.
Different types of high-energy ball mills have been developed for the HEBM
process, including the Spex vibrating mill, planetary ball mill, rotating ball mill,
attritors, and others.
42
In our nanotube synthesis research, two types of HEBM mills
have been used: a rotating ball mill and a vibrating-frame ball mill. A vibrating-
frame grinder (Pulverisette O, Fritsch) uses only one large ball (50 mm diameter).
The media of the ball and vial can be stainless steel or ceramic (tungsten carbide).
As shown in Fig. 7.3, the milling chamber has been modied, sealed with an O-
ring. The atmosphere can be changed via a valve and the pressure monitored with
an attached gauge.
43
The milling energy and frequency can be adjusted by varying
vibrating amplitude and ball mass. By measuring the balls movement, a milling
intensity is dened as
I = M
b
V
max
f/M
p
,
where M
b
is the mass of the milling ball, V
max
the maximum velocity of the vial, f
the impact frequency, and M
p
the mass of powder. The milling intensity is a very
important parameter to MA and reactive ball milling. For example, an intensity
threshold exists for full amorphization of Ni-Zr alloys.
44
When milling below this
TABLE 7.1
Comparison of Milling Parameters for Conventional Ball Milling and
High-Energy Ball Milling
Conventional Ball Milling High-Energy Ball Milling
Milling time <1 h 20200 h
Impact energy (W/g/ball) 0.001 0.2
Milling action Rolling Impact
Particle size m Sub-m (clusters)
Structural changes No Yes
Chemical reactions No Yes
Milling atmosphere No control Vacuum, gases
Milling temperature No control Liquid nitrogen 700C
threshold, a mixture of nanocrystalline and amorphous phases is produced. This
threshold depends on milling temperature and alloy composition.
Figure 7.4 shows a schematic of a steel rotating ball mill, consisting of a rotating
horizontal cell loaded with several hardened steel balls. As the cell rotates, the balls
drop onto the powder that is being ground. An external magnet is placed close to
the cell to increase milling energy.
45,46
Different milling actions can be realized by
adjusting the rotation rate and magnet position. The rolling balls give a shear
component of force to particles, while the falling balls provide head-on impacts with
high kinetic or milling energy.
47
FIGURE 7.3 A schematic of a modied vibrating ball mill (Fritsch, Pulverisette O), showing
impact actions and control of milling atmosphere and temperature.
FIGURE 7.4 A schematic of a high-energy rotating ball mill, showing milling actions.
Valve
O ring
Vial
Vibrating
frame
Elastic
band
Electromagnet
Thermacouple
Powder
Ball
During HEBM, material particles are repeatedly attened, fractured, and
rewelded. Every time two steel balls collide or one ball hits onto the chamber wall,
they trap some particles between their surfaces.
48
Such high-energy impacts
severely deform the particles and create atomically fresh, new surfaces, as well as
a high density of dislocations and other structural defects.
49
Depending on milling
conditions and reaction kinetics, mechanochemical reactions may be conned to
a very small volume during each impact due to an instantaneous injection of a
huge amount of mechanical energy into material particles.
50
Therefore, a high
number of impacts is required to complete a gradual transformation/reaction pro-
cess. The repeated fracturing and welding of powder particles increases the contact
areas between reactants and particles, allows fresh surfaces to come into contact
repeatedly, and allows the reaction often to proceed to completion without the
necessity for bulk diffusion of reactants through product layers.
51,52
Consequently,
reactions that normally require high temperatures can occur at lower temperatures
during HEBM. In addition, the high defect density induced by HEBM can accel-
erate the diffusion process.
53
Alternatively, the deformation and fracturing of par-
ticles cause continuous size reduction and can lead to reduction in diffusion
distances. This can at least reduce the reaction temperatures signicantly, even if
they do not occur at room temperature.
54,55
Since newly created surfaces are most
often very reactive and readily oxidize in air, the HEBM has to be conducted in
an inert atmosphere.
It is now recognized that high-energy ball milling, along with other nonequilib-
rium techniques such as rapid quenching, irradiation/ion implantation, plasma pro-
cessing, and gas deposition can produce a series of metastable and nanostructured
materials that are usually difcult to prepare using melting or conventional powder
metallurgy methods.
56,57
Detailed chemical reactions and phase transformations of
B and BN induced by HEBM are presented in the following sections.
7.2 SYNTHESIS OF BN NANOTUBES
FROM ELEMENTAL B
7.2.1 MILLING WITH A ROTATING STEEL MILL
Boron nitride nanotubes were prepared from elemental B powder via the following
two processes: high-energy ball milling and isothermal annealing. The milling exper-
iment was conducted using a steel rotating ball mill. The milling cell was rst loaded
with several grams of B powder and four steel balls (diameter of 25.4 mm). The mill
was then lled with reactive ammonia gas at a pressure of 300 kPa prior to milling.
The pressure inside the milling cell was monitored regularly using a pressure gauge.
The pressure changes recorded over the whole milling process are illustrated in
Fig. 7.5. A rapid reduction of pressure from 300 to 190 kPa occurred during the rst
36 h of milling, which was caused by absorption of ammonia gas onto fresh surfaces
of B particles newly created by high-energy impacts.
28
The particle size reduction is
the dominant event in the rst stage of milling. Interestingly the pressure inside the
mill started to rise during further milling and exceeded the starting pressure at the end
of 150 h (315 kPa). This pressure increase is indicative of the occurrence of nitriding
reactions between boron powder and the absorbed ammonia gas under HEBM. The
possible reaction is 2B + 2NH
3
> 2BN + 3H
2
. Hydrogen gas released from the
ammonia leads to a pressure rise inside the mill. Similar gassolid reactions induced
by HEBM were also observed in Ti(Zr)-NH
3
systems.
50,58,59
To conrm the preceding
reaction, chemical compositions of milled B samples were examined using x-ray
energy dispersive spectroscopy (EDS) in a scanning electron microscopy (SEM), as
well as combustion chemical analysis. Figure 7.6 shows the results obtained. The
FIGURE 7.5 Pressure changes inside the milling chamber of a steel rotating mill during
milling of B powder in NH
3
at room temperature.
28
FIGURE 7.6 Nitrogen content of milled B samples in the rotating mill as a function of milling
time.
29
100
150
200
250
300
350
400
0 20 40 60 80 100 120 140 160
P
r
e
s
s
u
r
e

(
k
P
a
)
Milling time (hours)
B+NH
3
BN + 3/2H
2
NH
3
absorption
0 20 40 60 80 100 120 140 160 180 200
0
2
4
6
8
10
12
14
N

(
a
t

%
)
gradual increase of the N content over the whole milling process is consistent with
ammonia absorption and a nitriding reaction that results in the formation of BN.
The high N content (10.1 wt%) and low H content (0.9 wt%) in the 150-h milled
sample conrm the presence of nitriding reactions during further milling stages.
HEBM also created signicant structural changes in B samples. The x-ray
diffraction (XRD) patterns of the sample after 150 h of milling in Fig. 7.7(a) shows
broadened peaks from unreacted B and a high background (a possible very large
peak at around 25) associated with amorphous B and newly formed nanostructural
BN phases. The new diffraction peak at around 51 was found to be associated with
Fe, which came from the steel milling balls and the cell as milling contamination.
TEM (Philips 430 and CM300 instrument equipped with an EDS system) analysis
of the milled B samples found a large number of Fe/Cr particles dispersed in a
disordered B matrix. Figure 7.8 shows that the Fe and Cr contents increase gradually
with increasing milling time and reached about 7 wt% for Fe and 1.4 wt% for Cr
at the end of milling.
29
These milling contaminations actually have a positive role
to play in nanotube formation. They might act as catalysts to help nanotube nucle-
ation. TEM did not nd any nanotubes, but rather thin layers of BN or B(N) solid
solution on the surface of B particles.
Isothermal annealing was conducted in a tube furnace under nitrogen gas ow
(0.5 l/min) so that the nitriding reaction could further proceed to full nitridation
with the formation of BN nanotubes. The B sample milled for 150 h was heated
at 1300C for 6 h. The XRD pattern in Fig. 7.7(b) shows pronounced peaks of
hexagonal BN and new diffraction peaks of an iron boride phase. No more B
peaks are found, suggesting that residual boron reacted completely with nitrogen
to form BN and some B formed Fe
x
B
y
. SEM observation found large aggregates
of nanosized laments and particles. Figure 7.9 shows a typical SEM image taken
from a sample milled for 150 h and heated at 1200C for 16 h in N
2
. Filaments
FIGURE 7.7 XRD patterns taken from (a) the 150-h milled B sample; (b) the milled sample
subsequently annealed at 1300C for 6 h in N
2
. O = hexagonal BN; + = Fe
x
B
y
.
28
20 40 60 80 100
O
O
O O
O
+
+
+
+
+
+
+
+
+
+
++
+
Boron (a)
(b)
D
i
f
f
r
a
c
t
i
o
n

i
n
t
e
n
s
i
t
y
(
a
r
b
i
t
r
a
r
y

u
n
i
t
s
)
2 (degrees)
with a diameter of 20 to 200 nm are observed. Thinner laments were difcult to
observe clearly due to the instrument resolution limit. The lament yield was
estimated by analyzing 25 SEM images taken from different aggregates. The ratios
of the area occupied by the laments to the whole sample area were calculated
for each image. The formation yield was found to be in the range of 65% to 85%.
29
TEM revealed that these laments have tubular structure. A typical TEM image
in Fig. 7.10 shows that thin (diameter less than 20 nm) nanotubes have typical
FIGURE 7.8 Iron and chromium contents of the milled B samples as a function of milling
time.
29
FIGURE 7.9 SEM image of a B sample rst milled for 150 h in NH
3
and then heated at
1200C for 16 h in N
2
.
0 20 40 60 80 100 120 140 160
0
1
2
3
4
5
6
7
8
Cr
Fe
W
t
.

%
parallel-walled, cylindrical structures. The parallel fringes on the wall areas of the
nanotube suggest eight atomic layers. Some thick nanotubes (diameter greater than
20 nm) exhibit a bamboo-like structure (Fig. 7.11). Nanosized polyhedral particles
of BN are also found. The chemical nature of these nanotubes was investigated using
electron energy loss spectroscopy (EELS) attached to a VG eld-emission transmis-
sion electron microscopy. In order to avoid carbon contamination, a BN nanotube
suspended over a hole in the carbon support lm was analyzed, and EELS spectra
are shown in Fig. 7.12. Two characteristic K-shell ionization edges (188 and 401
eV) of B and N are observed and the B/N ratio is 0.98. No carbon was detected from
any BN nanotubes or particles.
28
FIGURE 7.10 TEM image of two thin BN nanotubes with diameter less than 20 nm. Parallel
fringes suggest a multiwalled, cylindrical structure.
29
FIGURE 7.11 TEM image of BN nanotubes with bamboo-type structure.
The formation yield and sizes of nanotubes depend on both milling and annealing
conditions. For the rotating ball mill, extensive milling treatment up to 150 h under
high milling intensity seems to be required for a high formation yield. The yield
increases with increasing milling time, which might be due to higher N and Fe
contents in the prolonged milled samples, as shown in Fig. 7.6 and Fig. 7.8. The
long milling time can be reduced by using a fast (higher milling frequency) milling
device. The current rotating ball mill was designed for fundamental research pur-
poses and involved a low (about 14 Hz) milling frequency, so that the evolution of
chemical reactions or phase transformations induced by ball impacts could be exam-
ined over a reasonable milling time frame.
45,46
Annealing temperatures are critical to BN nanotube formation. A temperature
of 1000C appears to be too low for full nitridation and therefore the yield is very
low. The yield increases with increasing temperature up to 1300C. At 1400C, very
thick nanotubes of diameter more than 200 nm, as well as large BN particles, were
produced. By optimizing both milling and annealing conditions, a very high forma-
tion yield of about 80% can be achieved. Under controlled annealing conditions,
different types and sizes of nanotubes can be produced selectively. For example,
cylindrical nanotubes with a diameter less than 20 nm can be produced at a temper-
ature below 1200C; coarse bamboo or cone-structured tubes with diameter around
100 nm can be formed at temperatures above 1300C.
29
A large quantity up to 1 kg of BN nanotubes can be produced by running four
large-sized milling cells at the same time.
29
The results obtained show that the HEBM
and the presence of Fe particles each have important roles in nanotube formation.
To understand the effect of HEBM and roles of Fe in nanotube formation, a different
milling device (involving a different impact mode and using a ceramic vial and balls)
was tested.
FIGURE 7.12 EELS spectra taken from a BN nanotube suspended over a hole in a carbon
support lm showing K-shell ionization edges of B and N. The calculated B/N ratio is 0.98
0.16.
28
150 200 250 300 350 400 450
0
500
1000
1500
2000
B-K
N-K
Energy loss (eV)
P
h
o
t
o
d
i
o
d
e

c
o
u
n
t
s
7.2.2 MILLING WITH A VIBRATING CERAMIC MILL
As described previously, the vibrating ball mill uses only one large ball to hammer
crush powders at the bottom of the vial. To avoid metal contamination, a tungsten
carbide ball and vial were used. The high milling energy level (I = 2.17 ms
2
)
44
was
ensured using a heavy milling ball of 1 kg and a high vibrating amplitude of 1.5
mm. 2 grams of hexagonal B powder was milled in ammonia gas at a starting
pressure of 300 kPa for up to 168 h. The pressure decreased from 300 kPa to 266
kPa after 60 h of milling and increased up to 297 kPa after milling for 168 h. Again
a nitriding reaction was induced by high-energy ball milling.
16
The milled samples were heated at 1200C for different times up to 16 h. BN
nanotubes with very different structures were obtained. Very thin nanotubes with
only two or three atomic layers were found in the sample after heating for 10 h.
Bundles of the thin nanotubes formed (Fig. 7.13). Abundant thick-walled BN
nanotubes and some with conical structures were formed (Fig. 7.14(a)) in the sample
after annealing for 16 h. Some nanotubes contain a W metal particle at their tips,
implying different formation mechanisms to thin cylindrical tubes. The diameters
of such nanotubes are around 100 nm with lengths of several microns. As shown in
Fig. 7.14(b), the nanotubes have hundreds of crossed bands of BN, thus segmenting
each tube into many chambers and resembling the structure of bamboo. Clearly
annealing time has a very important effect on the nal nanotube structures.
HEBM with the ceramic vibrating mill created similar disordered structure in
B powder and the same nitriding reaction as the steel rotating mill. It conrms that
HEBM has an important role to play in BN nanotube formation rather than a simple
metal catalytic effect. No difference was found between pure impact milling actions
in the vibrating mill and the mixture of actions involving both shearing force and
impacts in the rotating mill.
FIGURE 7.13 TEM images showing a bundle of BN nanotubes, which was taken from a B
sample rst milled for 168 h in a ceramic vibrating mill and then heated at 1200C for 10 h in
N
2
. (Courtesy of Dr. John Fitzgerald.)
Although we did not expect any metal catalysts involved in the formation process,
W reduced from WC appears to act as a catalyst in the formation of many coarse
nanotubes, as seen in Fig. 7.14. The XRD patterns in Fig. 7.15 show that the milled
sample has a similar disordered structure to the B samples milled in a steel mill. The
weak WC peaks indicate the presence of milling contamination. The BN phase started
to form after 4 h of annealing. WC was rst reduced by B to W
2
B
5
. Some W
2
B
5
was
further reduced to W during longer heating for 10 and 16 h. An increased amount of
BN phase was produced for these longer times, as suggested by the more pronounced
peaks. The possible chemical reactions occurring during annealing are
WC + B > WxBy + C
WxBy + N > W + BN
FIGURE 7.14 TEM images show thick-walled BN nanotubes with conical structure. The
images were taken from a B sample rst milled for 168 h in a ceramic vibrating mill and
then heated at 1200C for 16 h in N
2
. (Image courtesy of Dr. John Fitzgerald.)
FIGURE 7.15 XRD patterns taken from samples: (a) B milled for 168 h in ceramic mill;
(bd) the milled sample subsequently annealed at 1200C in N
2
for various times marked.
16
20 40 60 80 100
H
x
x H
H
o
o o o
o
x
o
o
o
o
o
o o
o
oo o
o
o
o
o
o
H
x
x
x
x
x
c
c
c
c
c
o
x
x
x
x
x
+
+
+
H
H
(d) 16 h
(c) 10 h
(b) 4 h
(a) as-milled
D
i
f
f
r
a
c
t
i
o
n

i
n
t
e
n
s
i
t
y

(
a
r
b
i
t
r
a
r
y

u
n
i
t
)
H:BN, o:B, c:WC, +:W, x:W
2
B
5
Diffraction angle (2)
In conclusion, BN nanotubes were produced successfully from two different
milling devices and different milling media (stainless steel and ceramic tungsten
carbide). The nanotubes obtained have the same nanostructure as BN nanotubes
produced by other methods.
3,79
The aforementioned results suggest the following
possible formation sequences:
1. Reactive ball milling: B + NH
3
BN (nanoparticles/layers) + H
2
(gas)
Annealing: BN nanoparticles/layers BN nanotubes
2. Ball milling: B + NH
3
B(nanoparticles) + B(N) (nanoparticles) + H
2
(gas)
Annealing: B(nanoparticles) + N
2
BN(nanotubes)
In the rst formation sequence, reactive milling produces BN particles, which
transform to BN nanotubes during heating without further nitriding reaction. In the
second sequence, the nanosized, disordered B particles created by HEBM react with
N
2
during annealing and BN nanotubes are the end products of the reaction. The
second sequence is conrmed by ball milling of B in Ar gas (rather than NH
3
), so
the milled B contains nanosized B particles but not BN phase. BN nanotubes formed
after heating the milled sample at 1200C for 6 h in N
2
gas. To clarify the possibility
of the rst formation processes, a further experiment with BN compound as the
starting material for HEBM was carried out, and detailed results are introduced in
the following section.
7.3 SYNTHESIS OF BN NANOTUBES
FROM BN COMPOUNDS
Several grams of BN (purity > 98%) powder were milled in a steel rotating ball mill
with four hardened steel balls. To prevent oxidation contamination, the milling cell
was purged with nitrogen gas several times and a pressure of 300 kPa was established
prior to milling. Because a nitridation reaction was not required, the more reactive
ammonia gas was not used during the milling process in this case.
XRD analysis revealed that the starting hexagonal BN phase was transformed
to a highly disordered BN phase after milling for 140 h. The XRD pattern in Fig.
7.16(a) shows two weak and broadened peaks (referred as 1 and 2), which can
be associated with a turbostratic BN structure.
60
XRD patterns also show the presence
of iron from the milling balls and the cell. The overall Fe level measured by XEDS
is about 3.8 wt%. TEM did not nd any nanotubes but aggregates consisting of
particles below 50 nm, as shown in Fig. 7.17. The selected area diffraction pattern
shows three diffuse rings associated with (002), (100), and (110) planes with lattice
spacings of 3.61, 2.14, and 1.24 , respectively, which suggests a disordered nanoc-
rystalline BN structure. There may also be some amorphous BN present. The larger
(002) interlayer distance and the absence of a 3-D structure are features characteristic
of a turbostratic structure, which is consistent with the XRD results.
Following milling, nanosized BN powders were heated under N
2
gas ow (1 ml
min
1
) at 1200C for about 10 h. The N
2
gas was used to prevent oxidation during
the annealing rather than as a reactive gas. The TEM image in Fig. 7.18 shows the
presence of clusters of BN nanotubes as well as nanoparticles of outer diameter in
the range from 10 to 50 nm. These nanotubes have multiwalled, cylindrical struc-
tures. BN nanotubes with different tubular structures were produced at a higher
annealing temperature, as shown in the TEM image in Fig. 7.19, which was taken
from the sample annealed at 1300C for 10 h. It shows three bamboo-like tubes.
The biggest has a diameter of 280 nm and the smallest has a diameter of about 120
FIGURE 7.16 XRD patterns taken from sample: (a) BN milled for 140 h in N
2
in a steel
mill; (b) the milled sample subsequently annealed at 1200C for 10 h in N
2
. 1 and 2: mixture
of turbostratic and amorphous BN; + = hexagonal BN, = Fe; O = Fe(C) austenite alloy.
15
FIGURE 7.17 TEM image showing nanosized particles in the BN sample after milling for
140 h in a steel mill without heating treatment. The insert diffraction pattern suggests a
disordered turbostratic structure.
15
20 40 60 80 100
(a)
(b)
+
+
+
+
+
+
+
+
*
*
*
*
o
o
o
o
1
2
D
i
f
f
r
a
c
t
i
o
n

i
n
t
e
n
s
i
t
y
(
a
r
b
i
t
r
a
r
y

u
n
i
t
s
)
2 (degrees)
nm. Each tube contains a metal particle at its tip and EDS reveals that these particles
are mainly Fe with a little Ni and Cr. The outer diameter seems to be determined
by the overall size of the metal particles and the inner diameter is close to the
projected width of the edge of the Fe particle.
EELS analysis was conducted on a single nanotube. Figure 7.20 consists of a
bright eld (BF) image and three elemental maps of C, B, and N, which were
produced using EELS imaging with inelastic electrons of C, B, and N k-edges,
respectively. The BF image shows that a nanotube, containing a metal particle at its
tip, appears across a hole between a large aggregate of BN and the C support lm
of the TEM grid. The brighter the region in the elemental map, the larger the
elemental content. For the C map, compared with the bright C lm region, the darker
FIGURE 7.18 TEM image showing BN nanotubes and nanoparticles formed in the BN
samples after rst milling and then subsequently heated at 1200C for 10 h.
FIGURE 7.19 TEM image showing BN nanotubes with bamboo-like structures, which
formed in the BN samples after rst milling, and then subsequently heated at 1200C for
16 h. The dark particles at the tips are Fe.
15
contrast in the nanotube areas suggests absence of C in the nanotube. In contrast,
the B and N maps have light contrast only in the nanotube areas, clearly indicating
strong B and N content in the nanotube.
The formation of BN nanotubes from the milled BN compound conrms that the
rst reaction sequence proposed in Section 7.3 is possible. It is interesting to explore
how nanosized, disordered BN particles produced by HEBM are transformed into BN
tubules during annealing. Comparing the XRD patterns of the milled BN (Fig. 7.16(a))
with those of the heated sample (Fig. 7.16(b)), the pronounced diffraction peaks of
the BN in Fig. 7.16(b) indicate crystal growth of BN from the disordered BN during
the annealing. Indeed, many two-dimensional (002) layers, the lowest free-energy
layers for hexagonal BN,
61
were observed under TEM in the annealed samples. The
(002) layers appear to extend into various tubular morphologies, with or without the
catalytic effect of metal particles. At higher annealing temperatures (around 1300C),
the growth is faster and most tubes are both thicker and longer. It is interesting that
some cylindrical tubes appear to have grown without the presence of a contaminant
particle, while the bamboo-like tubes are most often capped with an Fe particle. Since
the starting phase is BN, no chemical reactions are likely to be involved in the
formation process of the nanotubes. BN vapor should not be present at such low
annealing temperatures. Therefore, a new solid-state nucleation and growth process
is indicated in the formation of BN, which is distinctly different from either the
previous case of reactive milling and annealing of B or the alternative vapor phase
(arc-discharge and laser ablation)
3,79
methods identied previously. Formation mech-
anisms are discussed in more detail in the following section.
FIGURE 7.20 EELS elemental maps of a BN nanotube lying across a hole between the BN
particle and the C support le. The rst BF image is a bright eld image. The brighter contrast
in the nanotube areas of the B and N maps suggest the nanotube observed is BN not C. (Image
courtesy of Dr. Jin Zou.)
7.4 FORMATION MECHANISM DISCUSSION
The previous sections demonstrate that ball milling and annealing produces a high-
yield of BN nanotubes with the same nanotubular structures as those produced by
other methods such as arc-discharge and laser ablation. How nanotubes are formed
in such a process and what formation mechanisms are involved are the questions to
be answered in this section.
7.4.1 SEPARATE NUCLEATION AND GROWTH
SEM and TEM observations found a high yield of BN nanotubes in heated samples
but not in any milled samples, suggesting that BN nanotubes develop only during
the thermal heating process following ball milling. What then is the role of high-
energy ball milling in the formation process? Can we produce BN nanotubes directly
from B or BN without the ball milling treatment? To examine this prospect, we
heated B and BN powders without milling treatment under the same annealing
conditions that produced nanotubes, but no nanotubes were obtained. This suggests
that the milling process has an essential role to play in nanotube formation during
subsequent annealing, which appears to be associated with the creation of suitable
nucleation sites for nanotube growth. We suggest, therefore, that the milling and
annealing processes actually correspond to separate nucleation and growth processes,
respectively, that is,
Ball milling > nucleation
Annealing > growth
The apparent separate nucleation and growth processes afford us the advantage of
investigating them separately.
7.4.2 NUCLEATION STRUCTURES
To explore possible nucleation structures, milled B and BN powders have been
carefully characterized using a number of techniques. In both milled B and BN, the
following common nanoscale structures were found:
1. Amorphous phase. Both XRD and TEM revealed that extended HEBM
leads to amorphization of both crystalline B and BN in both steel and
ceramic mills. As discussed previously, high-energy ball milling can trans-
form ordered crystalline structures to disordered amorphous phases above
certain milling intensity.
44
The milling intensities of the two mills used
appear to be sufciently high to create a predominantly amorphous struc-
ture in both B and BN, although a small amount of nanosized crystalline
material or layers less than 100 nm was still found after milling. The
amorphous phase is metastable and recrystallizes during heating.
2. Nanosized metal particles. A large number of metal particles were
introduced into B and BN powders during ball milling with both steel
and ceramic mills. The milling contamination is almost unavoidable by
the mechanical milling process, especially under high milling intensity.
In general, such milling contamination is undesirable, but, in the current
case, metal particles (Fe, Ni, Cr from the steel mill and W from the ceramic
mill) have been found to have positive catalytic roles in promoting nano
tube formation. Actually, a high yield of BN nanotubes can only be
obtained above a certain level of metal content (1 wt%). Such catalytic
effects have been reported extensively in the literature for nanotube for-
mations by other methods.
62,63
3. Microporous structures. Both milled B and BN powders have high surface
areas as a result of particle size reduction caused by mechanical grinding.
Figure 7.21 illustrates the surface area of BN samples as a function of
milling time. The total surface area of a BN sample has increased from 10
to 82 m
2
/g after milling for 150 h. Analysis by the t-plot method found
that almost half of the total surface area is actually contributed by
micropores.
64
For example, the 150-h milled sample has a micropore area
of 39.5 m
2
/g. This result shows that the milled particles contain a large
number of open micropores. Micropores presumably form during agglom-
eration and welding of BN particles under the ball impacts.
65
The nano-
porous structure also could be created by severe plastic deformation of the
(002) planes.
66
Indeed TEM analysis indicated a large number of severely
deformed (200) basal layers in milled samples.
The formation of BN nanotubes has two important requirements: the presence
of seeds or nucleation sites and the source of free atoms. Based on TEM observations,
we presume that possible seeds are both nanoscale metal particles and micropores
FIGURE 7.21 Variation of specic surface area of milled BN samples as a function of milling
time. The micropore area was determined using the T-plot method.
0 20 40 60 80 100 120 140 160
0
10
20
30
40
50
60
70
80
90
100
S
p
e
c
i
f
i
c

s
u
r
f
a
c
e

a
r
e
a

(
m
2
/
g
)
Total surface area
Micropore area
created by HEBM. Our experimental results have demonstrated clearly that Fe and
W particles can act as seeds to catalyze nanotube growth. In addition, our observations
suggest that micropores might also be responsible for the formation of nanotubes
without requiring the presence of metal particles. It is now reasonably well accepted
that multiwalled nanotubes can form without the help of metal particles, while the
formation of single-wall nanotubes appears to depend on specic metal catalysts.
67
Our data indicate that micropores containing four-membered squares might serve as
starting sites from which tubular structures are formed. Harris et al.
68
reported that
multiwalled C nanotubes can be formed by heating an arc-dischargeproduced carbon
soot containing fullerenes and micropores. A porous, amorphous phase and highly
disordered particles, observed in our ball milled samples, could be the source of free
atoms, which are available to facilitate nanotube growth since the amorphous and
disordered B or BN phases are metastable under heating. For example, XRD analysis
(Fig. 7.7(b), Fig. 7.15, and Fig. 7.16(b)) has revealed that the amorphous phase
crystallizes to form dominant (002) planes during heating. Therefore, milled B and
BN samples appear to contain at least two possible nucleation structures, as well as
a source of free atoms. These features make ball-milled material a perfect precursor
for nanotube growth.
7.4.3 THE GROWTH PROCESS
BN nanotubes grow out from BN particle clusters during isothermal annealing in a
temperature range of 11001300C. This temperature range is above the bulk self-
diffusion temperature of BN but well below the boiling point of B (2550C) and the
sublimation temperature of BN (3000C).
69
We therefore conclude that vapor phases
of B or BN are not possible during BN nanotube growth under our experimental
conditions. In our cases, both nucleation and growth processes seem to occur in the
solid state. At such a low growth temperature (12001400C), surface diffusion is
the most likely mechanism for nanotube growth.
70
Noncatalytic and metal-induced catalytic growth appear to be the basis of two
separate growth mechanisms. We now briey review our results and those of others
that affect such mechanisms.
1. Noncatalytic growth. Many BN nanotubes with perfect tip structures are
not associated with any metal particles. They appear to form without the
help of metal catalysts. For example, the TEM image in Fig. 7.22 shows
two BN nanotubes with only three walls and at tips. No metal particles
are found either inside the tubes or at the tips. Although a large metal
particle (dark spot) is present near the two tubes, it is completely covered
with thick BN layers, and should not be responsible for the formation of
the two small tubes. The small diameters of these nanotubes are close to
the micropore sizes. We suggest that micropores might be the nucleation
sites for these tubes. Indeed, carbon nanotubes have been successfully
produced from nanoporous carbon containing pentagons and
heptagons
7173
without using metal catalysts, and catalyst-free synthesis
of BN nanotubes has also been reported in a laser-ablation process.
13
Thus, solid-state crystallization is considered the main process for non-
catalytic growth in our case and surface diffusion may provide the driving
force for atomic motion that feeds material to the nanotube during growth.
2. Metal catalytic growth. Metal catalytic growth has been found in both B
(Fig. 7.11 and Fig. 7.14) and BN (Fig. 7.19 and Fig. 7.20) cases. The
popular metal-catalytic growth model for C nanotubes, especially for the
CVD process, is the vaporliquidsolid (VLS) mechanism.
62,63
Carbon
vapor dissolves into a metallic melt, forming a saturated solution at high
temperature; when the temperature is reduced or when temperature or
concentration gradients exist, carbon precipitates out in the form of nan-
otubes. Because of the lack of vapor phase in our case, the possible
dissolution and precipitation process that results in BN nanotubes is most
likely to be a solidliquidsolid process.
74,75
7.5 SUMMARY
BN nanotubes in large quantity and high formation yield have been prepared using
a ball millingannealing method consisting of high-energy ball milling of B or BN
powders and then subsequent annealing. The method has the advantages of control
in nanotube structures and sizes, high formation yield, and a much larger quantity
compared with other methods. The available large quantity of BN nanotubes makes
the studies of property and application possible. A stronger resistance to oxidation
of BN nanotubes has been found, which makes them a better choice than C nanotubes
for composite materials. The method can be scaled up to an industrial level and
therefore commercialization of BN nanotube production is possible. Systematic
research also found nanosized metal particles and micropores as possible nucleation
structures and the amorphous/disordered phase as the source of free B and N atoms.
FIGURE 7.22 TEM image showing two thin BN nanotubes without containing any metal
particles and an Fe particle covered by (002) basal planes nearby.
The growth process is via the solid state, driven by surface diffusion. In addition to
the synthesis of BN, C,
73,76
and other oxide
77,78
nanotubes, the ball milling technique
has also been used in processing of nanotubes, including opening of closed nano-
tubes,
7981
the cutting of long nanotubes into short lengths,
8184
and in assisting
functionalization,
85,86
helping aligned nanotube growth,
87,88
demonstrating the impor-
tant role of HEBM in nanotube research.
PROBLEMS
1. What are the structural differences between C and BN nanotubes?
2. Why do BN nanotubes have at tips instead of cone-shaped tips?
3. What are the different properties of BN nanotubes compared with C
nanotubes?
4. What is the role of high-energy ball milling in the milling-annealing process?
5. Is HEBM the same as conventional ball milling?
6. What are the possible applications of BN nanotubes based on their special
properties?
ACKNOWLEDGMENTS
The authors gratefully acknowledge the contributions from all collaborators, espe-
cially Professor Lewis Chadderton, Dr. John Fitzgerald, Dr. Jin Zou, Mr. Martin
Conway, Mrs. Jun Yu, Professor Stewart Campbell, and Professor Gerard Le Caer.
They also thank Electron Microscopy Units of the Australian National University
and of University of Sydney for their help with electron microscopes. This work
was supported nancially by several research grants from the Australian Research
Council.
REFERENCES
1. Iijima, S., Helical microtubules of graphitic carbon, Nature, 354, 56, 1991.
2. Rubio, A., Corkill, J.L., and Cohen, M.L., Theory of graphitic boron nitride nano-
tubes, Phys. Rev. B, 49, 5081, 1994.
3. Chopra, N.G. et al., Boron nitride nanotubes, Science, 269, 966, 1995.
4. Dresselhaus, M.S., Dresselhaus, G., and Eklund, P.C., Science of Fullerenes and
Carbon Nanotubes, Academic Press, New York, 1996, 756.
5. Ajayan, P.M., Carbon nanotubes: A new graphite architecture, Condensed Matter
News, 4, 9, 1995.
6. Chen, Y. et al., The nucleation and growth of carbon nanotubes in a mechano-thermal
process, Carbon, 42, 1543, 2004.
7. Loiseau, A. et al., Boron nitride nanotubes with reduced numbers of layers synthesized
by arc discharge, Phys. Rev. Lett., 76, 4737, 1996.
8. Terrones, M. et al., Metal particle catalysed production of nanoscale BN structures,
Chem. Phys. Lett., 259, 568, 1996.
9. Saito, Y., Maida, M., and Matsumoto, T., Structures of boron nitride nanotubes with
single-layer and multilayers produced by arc discharge, Jpn. J. Appl. Phys., 38, 159, 1999.
10. Menon, M., and Srivastava, D., Structure of boron nitride nanotubes: Tube closing
versus chirality, Chem. Phys. Lett., 307, 407, 1999.
11. Terrones, M. et al., Nanotubes: A revolution in materials science and electronics,
Fullerenes and Related Structures, 199, 189, 1999.
12. Golberg, D. et al., Fine structure of boron nitride nanotubes produced from carbon
nanotubes by a substitution reaction, J. Appl. Phys., 86, 2364, 1999.
13. Lee, R.S. et al., Catalyst-free synthesis of boron nitride single-wall nanotubes with
a preferred zig-zag conguration, Phys. Rev. B, 64, 1214051, 2001.
14. Endo, M., Iijima, S., and Dresselhaus, M.S., Carbon Nanotubes, 1st Ed., Elsevier
Science, Oxford, 1996, 27.
15. Chen, Y. et al., A solid state process for formation of boron nitride nanotubes, Appl.
Phys. Lett., 74, 2960, 1999.
16. Fitzgerald, J.D., Chen, Y., and Conway, M.J., Nanotube growth during annealing of
mechanically milled boron, Appl. Phys. A, 76, 107, 2003.
17. Chen, Y. et al., Mechanochemical synthesis of boron nitride nanotubes, Mater. Sci.
Forum, 312314, 173, 1999.
18. Bando, Y., Ogawa, K., and Golberg, D., Insulating nanocables: Invar Fe-Ni alloy
nanorods inside BN nanotubes, Chem. Phys. Lett., 347, 349, 2001.
19. Mickelson, W. et al., Packing C60 in boron nitride nanotubes, Science, 300, 467, 2003.
20. Chopra, N.G., and Zettl, A., Measurement of the elastic modulus of a multi-wall boron
nitride nanotube, Solid State Communications, 105, 297, 1998.
21. Blase X. et al., Stability and band gap constancy of boron nitride nanotubes, Europhys.
Lett., 28, 335, 1994.
22. Chen, Y., Zou, Y., Campbell, S.J., and Le Caer, G., Boron nitride nanotubes: Pro-
nounced resistance to oxidation, Appl. Phys. Lett., 84, 13, 2430, 2004.
23. Tsang, S.C., Harris, P.J.F., and Green, M.L.H., Thinning and opening of carbon
nanotubes by oxidation using carbon dioxide, Nature, 362, 520, 1993.
24. Ajayan, P.M., Opening carbon nanotubes with oxygen and implications for lling,
Nature, 362, 522, 1993.
25. Han, W. et al., Synthesis of boron nitride nanotubes from carbon nanotubes by a
substitution reaction, Appl. Phys. Lett., 73, 3085, 1998.
26. Han, W. et al., Transformation of BxCyNz nanotubes to pure BN nanotubes, Appl.
Phys. Lett., 81, 1110, 2002.
27. Cumings, J., and Zettl, A., Mass-production of boron nitride double-wall nanotubes
and nanococoons, Chem. Phys. Lett., 316, 211, 2000.
28. Chen, Y., Fitz Gerald, J., Williams, J.S., and Bulcock, S., Synthesis of boron nitride
nanotubes at low temperatures using reactive ball milling, Chem. Phys. Lett., 299,
260, 1999.
29. Chen, Y., Conway, M., Williams, J.S., and Zou, J., Large-quantity production of high-
yield boron nitride nanotubes, J. Mater. Res., 17, 8, 1896, 2002.
30. Benjamin, J.S., Dispersion strengthened superalloys by mechanical alloying, Metall.
Trans., 1, 2943, 1970.
31. Koch, C.C. et al., Preparation of amorphous Ni60Nb40 by mechanical alloying,
Appl. Phys. Lett., 43, 1017, 1983.
32. Suryanarayana, C., Nanocrystalline materials, Inter. Mater. Rev., 40, 41, 1995.
33. Debarbadillo, J.J., and Smith, G.D., Recent developments and challenges in the
application of mechanically alloyed, oxide dispersion strengthened alloys, Mater. Sci.
Forum, 8890, 167, 1992.
34. Chen, Y., Fitz Gerald, J., Chadderton, L.T., and Chaffron, L., Nanoporous carbon
produced by ball milling, Appl. Phys. Lett., 74, 19, 2782, 1999.
35. Tang, J. et al., Extended solubility and spin-glass behavior in a Ag-Gd solid solution
prepared by mechanical alloying, Phys. Rev. B, 52, 12829, 1995.
36. Koch, C.C., Amorphisation by mechanical alloying, J. Non-Crystalline Solids,
117118, 670, 1990.
37. Butyagin, P.J., The chemical forces in mechanical alloying, Mater. Sci. Forum, 8890,
695, 1992.
38. Schaffer, G.B., and McCormick, P.G., Displacement reactions during mechanical
alloying, Metall. Trans. A, 21, 2789, 1990.
39. Calka, A., and Williams, J.S., Synthesis of nitrides by mechanical alloying, Mater.
Sci. Forum, 8890, 787, 1992.
40. Chen, Y., and Williams, J.S., Formation of metal hydrides by mechanical alloying,
J. Alloys Compounds, 217, 181, 1995.
41. Matteazzi, P., and Le Caer, G., Room-temperature mechanosynthesis of carbides by
grinding of elemental powders, J. Am. Ceram. Soc., 74, 1382, 1991.
42. Suryanarayana, C., Mechanical alloying and milling, Prog. Mater. Sci., 46, 13, 2001.
43. Chen, Y., Le Hazif, R., and Martin, G., Amorphization in a vibrating frame grinder:
An example of phase transition in driven systems, Mater. Sci. Forum, 8890, 35, 1992.
44. Chen, Y. et al., Ball-milling-induced amorphization in Ni-Zr compounds: A parametric
study, Phys. Rev. B, 48, 14, 1993.
45. Chen, Y., Halstead, T., and Williams, J.S., Inuence of milling temperature and
atmosphere on the synthesis of iron nitrides by ball milling, Mater. Sci. Engin. A,
206, 24, 1996.
46. Calka, A., and Radlinski, R.P. Formation of amorphous Fe-B alloys by mechanical
alloying, Appl. Phys. Lett., 58, 119, 1991.
47. Calka, A. and Kikolov, J.I., The dynamics of magneto-ball milling and its effect on
phase transformations during mechanical alloying, Mater. Sci. Forum, 179181, 333,
1995.
48. Benjamin, J.S., and Volin, T.E., The mechanism of mechanical alloying, Metall.
Trans., 5, 1929, 1974.
49. Galy, D., Chaffron, L., and Martin, G., Amorphization mechanism of NiZr
2
by ball
milling, J. Mater. Res., 12, 688, 1997.
50. Chen, Y., and Williams, J.S., Investigation of gas-solid reactions realized by ball
milling, Mater. Sci. Forum, 225227, 545, 1996.
51. Suryanarayana, C., and Froes, F.H., Mechanical alloying of titanium-based alloys,
Adv. Mater., 5, 96, 1993.
52. Chen, Y. et al., Study on mechanism of mechanical activation, Mater. Sci. Engin. A,
226, 95, 1997.
53. Bhattacharya, A.K., and Arzt, E., Plastic deformation and its inuence on diffusion
process during mechanical alloying, Scripta Metall. Mater., 28, 395, 1993.
54. Chen, Y., Hwang, T., and Williams, J.S., Ball milling induced low-temperature car-
bothermic reduction of ilmenite, Mater. Lett., 28, 55, 1996.
55. Chen, Y. et al., Mechanically activated carbothermic reduction of ilmenite, Metall.
Trans. A, 28A, 1115, 1997.
56. Suryanarayana, C., Mechanical alloying and milling, Prog. Mater. Sci., 46, 1, 2001.
57. Gleiter, H., Nanostructured materials: State of the art and perspectives, Z. Metallkd.,
86, 78, 1995.
58. Chen, Y., Li, Z.L., and Williams, J.S., The evolution of hydriding and nitriding reactions
during ball milling of titanium in ammonia, J. Mater. Sci. Lett., 14, 542, 1995.
59. Chen, Y., and Williams, J.S., Competitive gas-solid reactions realized by ball milling
of Zr in ammonia gas, J. Mater. Res., 11, 6, 1500, 1996.
60. Hirano, S.I., Yogo, T., Asada, S., and Naka, S., Synthesis of amorphous boron nitride
by pressure pyrolysis of borazine, J. Am. Ceram. Soc., 72, 66, 1989.
61. Abrahamson, J., The surface energies of graphite, Carbon, 11, 337, 1973.
62. Tibbertts, G.G., Why are carbon laments tubular? J. Crystl. Growth, 66, 632, 1984.
63. Laurent, C. et al., Metal nanoparticles for the catalytic synthesis of carbon nanotubes,
New J. Chem., 22, 1229, 1998.
64. Seaton, W.G., A new analysis method for the determination of the pore size distribution
of porous carbon from nitrogen adsorption measurements, Carbon, 27, 853, 1989.
65. Berlin, O., Chemistry and Physics of Carbon, Marcel Dekker, New York, 1989, 94.
66. Huang, J.Y., Yasuda, H., and Mori, H., Highly curved carbon nanostructures produced
by ball-milling, Chem. Phys. Lett., 303, 130, 1999.
67. Charlier, J.C., and Iijima, S., Growth mechanisms of carbon nanotubes, in Dressehuas,
M.S., Dresselhaus, G., and Avouris, Ph., Carbon Nanotubes: Synthesis, Structure,
Properties, and Applications, Topics in Applied Physics 80, Springer-Verlag, New
York, 2001, 55.
68. Harris, P.J.F. et al., High-resolution electron microscopy studies of a microporous
carbon produced by arc-evaporation, J. Chem. Soc. Faraday Trans., 90, 2799, 1994.
69. West, R.C. et al., Handbook of Chemistry and Physics, 70th Ed., CRC Press, Boca
Raton, Florida, 1990, B77.
70. Chadderton, L.T., and Chen, Y., Nanotube growth by surface diffusion, Phys. Lett.
A, 263, 401, 1999.
71. Iijima, S., Growth of carbon nanotubes, Mater. Sci. Eng., B19, 172, 1993.
72. Setlur, A.A. et al., A promising pathway to make multiwalled carbon nanotubes, Appl.
Phys. Lett., 76, 3008, 2000.
73. Chen, Y., Conway, M.J., and Fitzgerald, J.D., Carbon nanotubes formed in graphite
after mechanically grinding and thermal annealing, Appl. Phys. A, 76, 633, 2003.
74. Chen, Y. et al., Solid-state formation of carbon and boron nitride nanotubes, Mater.
Sci. Forum, 343, 63, 2000.
75. Chadderton, L.T., and Chen, Y., A model for the growth of bamboo and skeletal
nanotubes: Catalytic capillarity, J. Crystl. Growth, 240, 164, 2002.
76. Chen, Y. et al., Investigation of nanoporous carbon powders produced by high energy
ball milling and formation of carbon nanotubes during subsequent annealing, Mater.
Sci. Forum, 312314, 375, 1999.
77. Kholmanov, I.N. et al., A simple method for the synthesis of silicon carbide nanorods,
J. Nanosci. Nanotechnol., 2, 5, 453, 2002.
78. Li, S.L. et al., Synthesis of TiTe2 nanotubes, Acta Chimica Sinica., 62, 6, 634, 2004.
79. Li, Y.B. et al., Transformation of carbon nanotubes to nanoparticles by ball milling
process, Carbon, 37, 3, 493, 1999.
80. Gao, B. et al., Electrochemical intercalation of single-walled carbon nanotubes with
lithium, Chem. Phys. Lett., 307, 153, 1999.
81. Pierard, N. et al., Production of short carbon nanotubes with open tips by ball milling,
Chem. Phys. Lett., 335, 1, 2001.
82. Kim, Y.A. et al., Effect of ball milling on morphology of cup-stacked carbon nano-
tubes, Chem. Phys. Lett., 355, 279, 2002.
83. Shimoda, H. et al., Lithium intercalation into opened single-wall carbon nanotubes:
Storage capacity and electronic properties, Phys. Rev. Lett., 88, 155021, 2002.
84. Liu. F. et al., Preparation of short carbon nanotubes by mechanical ball milling and
their hydrogen adsorption behavior, Carbon, 41, 2527, 2003.
85. Breton, Y. et al., Functionalization of multiwall carbon nanotubes: Properties of
nanotubes-epoxy composites, Molecular Crystals and Liquid Crystals, 387, 359, 2002.
86. Wang, Y., Wu. J., and Wei, F., A treatment method to give separated multi-walled
carbon nanotubes with high purity, high crystallization and a large aspect ratio,
Carbon, 41, 15, 2939, 2003.
87. Guo, X.M., Qi, J.F., and Sakurai, K., Mechanochemical formation of novel catalyst
for preparing carbon nanotubes: Nanocrystalline yttrium aluminum iron perovskite,
Scripta Materialia, 48, 8, 1185, 2003.
88. Chen, Y., and Chadderton, L.T., Improved growth of aligned carbon nanotubes by
ball milling, J. Mater. Res., 19, 10, 2791, 2004.
8
Plasma Deposition
of Ultra-Thin Functional
Films on Nanoscale
Materials
Peng He and Donglu Shi
CONTENTS
8.1 Introduction
8.2 The Plasma-Coating Technique
8.3 Applications and Characterization
8.3.1 Plasma Polymer Films Deposited on ZnO Particles
8.3.2 Plasma Coating of Magnetic Nanoparticles
8.3.3 Copolymer Coating of YYbErO
2
S
8.3.4 Plasma Coating of Carbon Nanofibers
8.3.5 Plasma Coating of an Aligned Carbon Nanotube Array
8.4 Processing and Characterization of Nanocomposite Materials
8.5 Summary
Problems
References
8.1 INTRODUCTION
Nanotechnology is the application of nanoscale materials. This has been the most
important research topic in the world for the past few years, and it will change our
lives in the very near future. One of the critical aspects of nanotechnology research
is how to modify the surface of nanoparticles to make them more useful for different
applications. This chapter describes a novel plasma polymerization technique for
coating nanoparticles, nanotubes, and nanoplates. The coated nanoparticles/tubes/
plates can be used to reinforce polymer materials and to provide special functional
capabilities for materials.
In the current development of nanomaterials, it has become critical to modify
the surfaces of the nanoparticles for both fundamental research and engineering
applications.
18
Taking different film technologies into consideration, plasma
polymerization coating as well as plasma surface treatment of powders are partic-
ularly promising because of the following unique features and advantages: (1) The
starting feed gases used may not contain the type of functional groups, such as
carbon double bond or triple bond, that are normally associated with conventional
polymerization; (2) such films are often highly cohesive and adhesive to a variety
of substrates, including conventional polymers, glasses, and metals; (3) polymer-
ization may be achieved without the use of solvents; (4) plasma polymer films can
be easily produced with thicknesses from several nanometers to 1 m; and (5)
through careful control of the polymerization parameters, it is possible to tailor the
films with specific chemical functionality, thickness, and other chemical and phys-
ical properties.
Plasma-treated powders have a wide range of potential applications. The appli-
cations involve improving the following properties of powders: biocompatibility,
sorption, wettability, triboelectricity, barrier and insulation, adhesion, diffusion, fric-
tion coefficient, corrosion, dispersion, flowability, and other properties. Plasma treat-
ment has wide applications in medicine, pharmacy, the pigment industry, rubber
chemistry, and others. The current research on plasma treatment of particles is mainly
focused on the micron size or even larger. In the literature, not much research has
been done investigating thin-film coating of nanosize particles. The long term main
objective of the research reported here is to develop a broad set of techniques for
plasma thin-film coating of nanoparticles and nanotubes. In order to develop this
coating technology, different monomers such as styrene, pyrrole, and C
6
F
14
, and
different substrates such as nanoparticles and carbon nanotubes have been used to
form new material systems. The processing and characterization of these materials
is discussed next.
8.2 THE PLASMA-COATING TECHNIQUE
The plasma-coating facility is a homemade system. The schematic diagram of the
plasma reactor for thin-film deposition of nanoparticles is shown in Fig. 8.1. It
consists mainly of a radio frequency (RF) source, a glass vacuum chamber, and
pressure gauge. The vacuum chamber of the plasma reactor has a long Pyrex glass
column about 80 cm in height and 6 cm in internal diameter.
911
The powder is
vigorously stirred at the bottom of the tube and thus the surface of particles can
be continuously renewed and exposed to the plasma for thin-film deposition during
the plasma polymerization processing. A magnetic bar is used to stir the powders.
The gases and monomers are introduced from the gas inlet during the plasma
cleaning treatment and plasma polymerization. Before the plasma treatment, the
basic pressure is pumped down to less than 50 mtorr and then the carrier gas (such
as argon) or monomer vapors are introduced into the reactor chamber. The operating
pressure is adjusted by the gas/monomer mass flow rate. During the plasma poly-
merization processing, the input power is 1080 W, and the system pressure is
300450 mtorr. The plasma treatment time varies from 15 to 120 min according
to the different monomers used and the different film thicknesses desired.
After the plasma treatment, the treated powders are examined by using trans-
mission electron microscopy (TEM), secondary ion mass spectrometry (TOFSIMS),
and Fourier transform infrared (FTIR). The high-resolution TEM (HRTEM) exper-
iments are performed on a JEOL JEM 4000EX TEM.
In this research, the substrates used for plasma deposition include ZnO, NiFe
2
O
4
,
YYbErO
2
S, and carbon nanofibers. The ZnO powders are obtained from Zinc Cor-
poration of America (Monaca, PA). The NiFe
2
O
4
is obtained from Inframat Corpo-
ration (Willington, CT). The YYbErO
2
S is obtained from OraSure Technologies,
Inc. (Bethlehem, PA). The carbon nanofiber Pyrograf-III PR-24 is obtained from
Applied Science, Inc., (Cedarville, OH). The monomers, such as pyrrole, acrylic
acid, C
6
F
14
, and styrene, were obtained from Alfa Aesar, a Johnson Matthey Com-
pany (Ward Hill, MA).
8.3 APPLICATIONS AND CHARACTERIZATION
Different applications of the plasma technique and the methods used for character-
izing the new material systems produced are described below.
8.3.1 PLASMA POLYMER FILMS DEPOSITED ON ZnO PARTICLES
In this experiment, a plasma polymer film is deposited on ZnO particles.
12
In trans-
mission electron microscopy (TEM), the original and coated ZnO nanoparticles were
dispersed onto a holy-carbon film supported by Cu grids for the TEM operated at
400 kV. Figure 8.2 shows the HRTEM image of the original, uncoated ZnO particles.
As can be seen in this figure, the particle sizes are in the several hundreds of
nanometers range. The particles also exhibit nonspherical shapes.
Figure 8.3 shows the HRTEM images of the coated particles. Compared with
Fig. 8.2, a bright ring was found around the particles. From the HRTEM images,
the bright ring was found to be one kind of amorphous layer outside the lattice
FIGURE 8.1 The plasma reactor for thin-film coating of nanoparticles: (a) schematic dia-
gram, (b) fluidized bed reactor.
Clamp
Clamp
15 cm
50 cm
15 cm
16 cm
Vacuum pump
Powder
Stirrer
6 cm
Pressure gauge
Monomer +
carrier gas inlet
R
F

+

i
m
p
l
o
s
i
o
n

s
h
i
e
l
d
(b) (a)
RF generator
40 cm
structure. The ZnO particles have a crystalline structure. Thus the lattice structure
in the images is the ZnO particle substrate. The amorphous layer that does not appear
in the uncoated images (Fig. 8.2) is due to the coating. In Fig. 8.3, the low magni-
fication images show a uniform coating on the particle surfaces. The coating thick-
ness is approximately 1020 nm thick over the entire particle surface. Although the
shape of particles is nonspherical and at some places the shape is even very sharp,
it is particularly interesting that the coating remains the same thickness, indicating
that the plasma chamber produces a uniform coating on all the particles.
To confirm the TEM observations shown in Fig. 8.2 and Fig. 8.3, TOFSIMS
was carried out to study the surface films of the particles. Figure 8.4(a) and (b) show
the positive and negative TOFSIMS spectra of the coated ZnO particles. In
Fig. 8.4(a), one can see that the spectrum of the positive ion from the coated ZnO
have strong peaks of functional groups such as C
4
H
7
+
, C
4
H
9
+
, C
6
H
13
O
4
+
, C
7
H
9
COH
+
,
and C
7
H
9
COOH
+
, indicating a surface coating on the particles consistent with the
HRTEM data presented in Fig. 8.2 and Fig. 8.3. In Fig. 8.4(b), where the spectra of
the negative ions are presented, we also see the AA monomer, AA dimer, and AA
dimer + C
2
H
4
and AA dimer + C
3
H
6
, which are strong indications of these functional
groups. These are the typical characteristic cluster patterns of a plasma-polymerized
poly(acrylic acid) film.
FTIR was used to study the effect of the plasma power on the molecular structure
of the AA films. As can be seen in Fig. 8.5(a), the strong C O peak near 1700 cm
1
FIGURE 8.2 TEM image of uncoated ZnO at different resolutions (a), (b), (c), and (d).
indicates the surface coating of the nanoparticles, and it is consistent with the
HRTEM data presented in Fig. 8.3. The peak of C O intensifies as the plasma
power increases up to 80 W, indicating a strong plasma power dependence. The
intensification of this peak also indicates that an increased number of C O function
groups came from the acrylic acid.
In order to investigate the solubility of the plasma film, the coated ZnO powder
was immersed into Ni
+
solution. After 1 h, the sample was removed from the solution,
dried, and the FTIR experiment was performed again on the nickel solution treated
powder. The results are shown in Fig. 8.5(b), (c), and (d). In these figure parts, we
can see that the AA film was entirely dissolved when the plasma power was 15 W.
The FTIR spectra clearly show that the film was removed by the nickel solution
after immersed for only 1 h. However, as the plasma power was increased to 60 W
(Fig. 8.5(c)) and 80 W (Fig. 8.5(d)), the AA film remained after immersed in the
nickel solution. This indicates that large cross links exist in the polymer, which
prevented the AA film from dissolving in the nickel solution; even the polymer has
the strong hydrophilic function groups COOH. Modified ZnO powder can be used
in epoxy, PE, PP, or PS to improve the mechanical properties. At the same time, the
ZnO powder can introduce special optical properties to composite materials.
FIGURE 8.3 Bright-field HRTEM images of the AA-coated ZnO nanoparticles: (a) at low
magnification; (b) at higher magnification; (c, d) showing the amorphous nature of the AA
thin film and the crystal lattices of the ZnO structure.
FIGURE 8.4 SIMS spectra of coated samples: (a) positive image (positive species); (b) nega-
tive image (negative species).
2.5
2.0
1.5
1.0
0.5
0.0
10
4
1.5
1.0
0.5
0.0
0.0
10
4
10
3
10
2
10
3
1.6
1.2
0.8
0.4
5.0
4.0
3.0
2.0
1.0
0.0
1.0
0.8
0.4
2.0
0.0
0.6
10
5
6.0
5.0
3.0
2.0
1.0
4.0
0.0
10
4
2.5
2.0
1.0
0.5
1.5
0.0
10
3
5.0
4.0
2.0
1.0
3.0
0.0
10
3
3.0
2.5
1.5
1.0
0.5
2.0
0.0
10
5
1.2
0.8
0.4
0.0
170 180 190 Mass 200
130 140 150 Mass
Mass
Mass
160
90 100 110 120
50 60 70 80
Mass 10 20 30 40
Mass 200 170 180 190
Mass 160 130 140 150
Mass 120 90 100 110
Mass
80 50 40 60 70
Mass 40 10 0 20 30
Na
C
2
H
3
+
C
2
H
5
+
C
3
H
5
+
C
3
H
5
+
C
4
H
7
+
C
7
H
7
+
C
6
H
13
O
4
C
7
H
9
COH
+
C
7
H
9
COOH
+
COOH
CH
3
COO
C
2
H
5
COO
C
2
H
3
COO
SO
4
H
C
4
H
9
+
Zn
+
Zn
2
+
ZnO
+
, ZnOH
+
COOH
+
O
OH
Cl
SO
4
AA dimer
AA dimer + C
2
H
4
AA dimer + C
3
H
6
AA dimer
+ C
4
H
8
Mass, amu
I
n
t
e
n
s
i
t
y
,

a
.
u
.
I
n
t
e
n
s
i
t
y
,

a
.
u
.
(a)
(b)
8.3.2 PLASMA COATING OF MAGNETIC NANOPARTICLES
In this experiment, NiFe
2
O
4
particles were coated using mechanical stirring instead
of magnetic stirring because of the particles magnetic properties. Figure 8.6 shows
the HRTEM image of the original, uncoated NiFe
2
O
4
particles. As can be seen in
this figure, the particle sizes range from 1050 nm, and the particles are severely
aggregated. Figure 8.6(c) is the high-resolution image of the original NiFe
2
O
4
nano-
particles. The lattice image further reveals the crystallographic features of the naked
NiFe
2
O
4
particles.
FIGURE 8.5 (a) FTIR spectra of a AA-coated ZnO at different plasma powers. The FTIR
spectra of Ni solution treated, AA-coated ZnO nanoparticles at plasma power of (b) 15 W,
(c) 60 W, and (d) 80 W.
FIGURE 8.6 HRTEM images of the original NiFe
2
O
4
nanoparticles (a, b) at different reso-
lutions, with a particle size distribution ranging from 10 to 50 nm, and (c) crystal lattice and
the uncoated nature of the nanoparticle surfaces.
1800 1700 1600 1500 1400 1200 1300 1100 1000
T
r
a
n
s
m
i
t
t
a
n
c
e
1800 1700 1600 1500 1400 1200 1300 1100 1000
T
r
a
n
s
m
i
t
t
a
n
c
e
Wavenumber (m
1
) Wavenumber (m
1
)
Power 60 W
Power 80 W
1800 1700 1600 1500 1400 1200 1300 1100 1000
T
r
a
n
s
m
i
t
t
a
n
c
e
1800 1700 1600 1500 1400 1200 1300 1100 1000
T
r
a
n
s
m
i
t
t
a
n
c
e
Wavenumber (m
1
) Wavenumber (m
1
)
After treatment
After treatment
Before treatment
Before treatment
After treatment
Before treatment
Power 15 W
(a)
(b)
(c) (d)
Power 30 W
From Fig. 8.7, we can see the magnetic particles are still aggregated. The coating
is only on the outside of the cluster, not exactly on the surface of each particle.
It is noted that the impurity particles such as TiOx and Cu (Fig. 8.8) have a very
uniform coating on each individual particle. We can reach the conclusion that the
mechanical stirring is satisfactory for many nanoparticles, but not for the magnetic
particles. The natural magnetic properties of NiFe
2
O
4
make them much more difficult
to disperse than regular particles.
In Fig. 8.9, the spectrum exhibits several very strong peaks between 1400 cm
1
and 1600 cm
1
, which is due to the C C vibrating special absorption of the benzene
ring. These peaks confirm that the plasma polymer structure contains some benzene
rings. In the FTIR absorption spectrum, the range from 690 cm
1
to 900 cm
1
belongs
to the benzene C-H out of plane bending.
In Fig. 8.9, at 758 cm
1
and 700 cm
1
there are two peaks, which are the special
absorption for one hydrogen atom on a benzene ring substituted by other function groups.
Figure 8.9 is more complicated than the regular FTIR spectra of polystyrene, but the
special absorption peaks for styrene are almost the same as those of regular polystyrene.
This means that in plasma polymerization the molecular structure is destroyed by ion or
electron bombardment, but the coating film still keeps some of the polystyrene molecular
structure, which can be used to get good compatibility with a polystyrene matrix.
FIGURE 8.7 TEM image of the coated NiFe
2
O
4
at different positions (a), (b), (c), and (d).
8.3.3 COPOLYMER COATING OF YYbErO
2
S
After using a single monomer to coat nanoparticles, several monomers are used
together to modify the surface properties of nanoparticles. Different monomers can
be used to introduce different function groups, which enable us to easily control the
surface properties of nanoparticles and to easily modify the surface properties of
nanoparticles for different purposes.
The particle used here is one type of luminescent particle. This particle will be
used as the substrate for bacteria detection. Other applications with different coatings
might be to use the nanoparticles in composites to detect cracks in structural mate-
rials, for radiation shielding, and other applications. For the current application, the
outside coating needs to contain the C O function group, which can be used to
attract the bacteria. However, the concentration needs to be carefully controlled for
different conditions. In order to get the C O function group and control the
FIGURE 8.8 TEM image of coated impurity: (a) Cu; (b) TiOx in magnetic particles.
concentration, two kinds of monomers are used in this coating processing. One is
methyl methacrylate (MMA), which used to introduce the C O function group.
Another one is styrene, which is used to control the concentration of C O.
Figure 8.10 is the HRTEM image of the coated particle. In Fig. 8.10, there is
a bright ring on the particle surfaces, which is the result of the polymer coating.
Figure 8.10 contains high magnification images showing a uniform coating on the
particle surfaces. The coating thickness is approximately 10 nm over the entire particle
surface. The coating layer is amorphous based on the HRTEMs of different particles.
Figure 8.11 shows there are several strong peaks belonging to different function
groups. The peak around 1700 cm
1
is the special absorption of C O vibration. The
peaks between 1450 cm
1
and 1700 cm
1
belong to the C-C vibration of the benzene ring.
From the molecular structure of the monomer, we can see that the benzene ring
comes from styrene and the C O function group comes from the MMA. Comparing
Fig. 8.11(a), (b), and (c), the intensity of the peaks between 1450 and 1600 cm
1
and
the peaks around 700 cm
1
and 760 cm
1
decreases as the ratio of styrene/MMA
decreases, and the intensity of peaks changing means the concentration of different
function groups has changed. With the MMA concentration increasing, the peak
around 1700 cm
1
becomes stronger and stronger. This evidence means that as more
MMA monomer is used in plasma coating, more C O function groups occur in the
coating films. The FTIR data strongly prove that through controlling the ratio of
monomers, the molecular structure of the coating polymer can be modified as desired.
The molecular structure of the monomers styrene and methyl methacrylate
(MMA) are shown in Fig. 8.12.
8.3.4 PLASMA COATING OF CARBON NANOFIBERS
After modifying nanoparticles, the so-called zero dimension substrate, the plasma-
coating technique is applied to one-dimensional substrates. Commercial Pyrograf
FIGURE 8.9 FTIR spectra of the coating of NiFe
2
O
4
nanoparticles.
4000 3500 3000 2500 2000 1500 1000 500
Wavenumber (cm
1
)
0
10
20
30
40
50
60
70
80
T
r
a
n
s
m
i
t
t
a
n
c
e
3
0
2
7
.
7
2
5
3
1
.
4
9
1
6
0
1
.
9
1
4
1
4
.
5
4
4
5
2
.
9
4
1
3
7
5
.
9
3
1
0
7
5
.
5
6
1
0
2
9
.
3
7
5
8
.
6
8
7
0
0
.
6
8
3
III carbon nanofibers were chosen as the substrate in this experiment.
13
The
nanofibers are stirred in the sample container at the middle of the chamber, and
thus the surface of the particles can be continuously renewed and exposed to the
plasma for thin-film deposition during the plasma polymerization process. Two
small magnetic bars were used to stir the powder. Because the density of the
nanofiber is very small, a mesh cover was used to prove the nanofibers spray out
from the sample container during the vigorous stirring. Before the plasma coating,
the argon gas was introduced for a surface cleaning processing. After 30 min of
ion bombardment, styrene and the C
6
F
14
gas with a controlled ratio were introduced
from another gas inlet. At the same time, the argon gas was shut off. The operating
pressure was adjusted by the gas/monomer mass flow rate. The total pressure was
about 350 mtorr. During the plasma polymerization process, the input power was
FIGURE 8.10 Bright-field HRTEM images of the coated YYbErO
2
S: (a) image of the original,
uncoated YYbErO
2
S nanoparticles with a particle size distribution ranging from 400 to 600
nm; (b), (c) image of the plasma coated YYbErO
2
S nanoparticles; (d) nanoparticles showing
crystal lattice and the amorphous nanoparticle surfaces.
15 W. The time of plasma treatment was 10 min. Per batch, 0.1 g of powder were
treated.
Figure 8.13 shows the bright-field TEM images of the original, uncoated Pyrograf
III PR-24 carbon nanofibers. As can be seen in this figure, the carbon nanofibers
have multiple walls with similar features to a multiwall carbon nanotube. However,
the nanofibers have the tubes at a 20 angle to the fiber axis and thus the tubes
terminate as shown in Fig. 8.14. The Pyrograf III PR-24 nanofibers have outside
diameters averaging about 70 nm and are quite uniformly distributed. Some nanofi-
bers become curved during their growth with the ends open. The HRTEM image
(Fig. 8.13(a)) of the original Pyrograf III PR-24 carbon nanofibers shows the graphite
structure with the interlayer spacing d
002
0.34 nm. Based on the bright-field TEM
and HRTEM images, the wall thickness of the nanofibers can be estimated to be
about 2030 nm. Nanofibers with axially parallel graphite layers and the nanofibers
with axially parallel graphite layers oriented at an angle to the tube axis (Fig. 8.13(a))
are observed. The edge dislocations can be seen due to the disorder of the graphite
layers (002). It can be seen that both outer and inner surfaces terminate at the graphite
(002) layer without an addition of surface layer (the edge is very sharp) for originally
uncoated nanofibers (Fig. 8.13(a)). The bright-field and high-resolution TEM images
FIGURE 8.11 FTIR of the coated particle (a) condition: Styrene 200 mtorr, MMA 50 mtorr,
(b) condition: styrene 150 mtorr, MMA 100 mtorr, (c) condition: styrene 100 mtorr, MMA
150 mtorr.
FIGURE 8.12 Molecular structure of the styrene and methyl methacrylate (MMA) monomers.
2000 1600 1200 800 400 0
T
r
a
n
s
m
i
t
t
a
n
c
e
7
0
1
.
5
8
7
6
0
.
1
8
1
3
7
7
.
3
1
4
5
2
.
5
1
4
9
5
1
6
0
1
.
2
1
7
1
2
.
6
(a)
Wavenumber (cm
1
)
2000 1600 1200 800 400 0
T
r
a
n
s
m
i
t
t
a
n
c
e
1
7
1
8
1
6
0
3
1
4
5
4
.
5
1
4
5
6
.
9
1
3
7
9
.
2
7
6
3
.
9
7
0
2
.
1
1
(b)
Wavenumber (cm
1
)
2000 1600 1200 800 400 0
T
r
a
n
s
m
i
t
t
a
n
c
e
1
7
1
4
.
8
1
6
0
8
.
8
1
4
5
4
.
5
1
3
7
7
.
3
7
6
2
.
4
7
0
4
.
1
(c)
Wavenumber (cm
1
)
Stryrene Methyl methacrylate
of these nanofibers after plasma treatment are shown in Fig. 8.14(b) and Fig. 8.15,
respectively. An ultrathin-film amorphous layer can be clearly seen over the surfaces
of the Pyrograf III PR-24 nanofibers (Fig. 8.14(b)). The thin film is uniform on the
surfaces of the nanofiber (Fig. 8.14(b)). The thickness of ultrathin film is approxi-
mately 27 nm all the way surrounding the nanofiber surfaces and is thicker and
more uniform than the disturbance (1 nm) on the outer surface of carbon nanofibers
FIGURE 8.13 Bright-field TEM images of the original, uncoated carbon nanofibers.
FIGURE 8.14 HRTEM images of nanofibers: (a) the fragments of the wall with inclined
planes (002) showing lattice space on the outer and inner surfaces of uncoated nanofibers
with slight disturbance (1 nm) on the surface; (b) an ultrathin film can be observed on the
surface of coated nanofibers.
200 nm
(a)
200 nm
(b)
(a)
Outer surface
0.34 nm
5 nm
(b)
10 nm
Outer coating ~7 nm
Inner coating 1~3 nm
(Fig. 8.14(a)). This coating can only be attributed to be a remarkably uniform layer
of the plasma coating. In Fig. 8.14(b), we show the HRTEM image of a coated
Pyrograf III PR-24 carbon nanofiber. The lattice image of graphite can be clearly
seen with an extremely thin layer of polymer film on its surface.
To confirm the TEM observations shown in Fig. 8.14 and Fig. 8.15, TOFSIMS
was carried out to study the surface films of the nanofibers. Figure 8.16 and Fig. 8.17
show the positive TOFSIMS spectra of coated and uncoated nanofibers. In Fig. 8.17,
one can see that the spectra of the positive ion from the uncoated nanofibers has strong
peaks of functional groups such as C
1
, C
2
, C
3
, C
4
, C
7
H
7
and C
10
H
8
, indicating that
the surface of nanofibers contains a hydrocarbon. Because the plasma coating of
polystyrene contains only carbon and hydrogen, the hydrocarbon from the surface of
nanofibers will cause a problem when trying to identify the coating from the nanofibers
surface. In order to solve this problem, a small amount of C
6
F
14
was added to copo-
lymerize with the styrene monomer. In Fig. 18, one can see that the spectrum of the
positive ion from the coated nanofibers has strong peaks such as CF
, C
2
F
, C
4
F
6
,
C
3
F
7
, C
4
F
7
, and C
5
F
7
, indicating fluorine in the coating film. The fluorine can come

only from the monomer, which strongly proves the existence of the coated film.
8.3.5 PLASMA COATING OF AN ALIGNED CARBON
NANOTUBE ARRAY
Compared with other CNT materials, aligned carbon nanotube array (ACNT) has better
organization and orientation. It will be easier to use in application areas. But CNT is
FIGURE 8.15 The bright-field TEM (a) and HRTEM images (b) showing the coating on the
surface of the nanofibers. The lattice image of carbon can be clearly seen with an extremely
thin layer of polymer film (a2 nm) on the outer surface of the coated Pyrograf III PR-24
carbon nanofiber.
inert toward most chemicals at room temperature. In order to modify the surface,
certain very strong dangerous solutions and chemicals, such as concentrated nitric acid
or concentrated sulfuric acid, must be used. The whole processing of surface modifi-
cation will last about 1020 h. Especially for ACNT, strong chemicals will not only
modify the surface of the CNT, but also damage its unique forest structure. In order
to find a better way to solve this problem, the plasma modification method was chosen.
Compared with the nanofibers, ACNT is more suitable for plasma treatment. All the
tubes grow along the same direction on the two-dimensional flat substrate. The gap
between each individual tube can let the monomer gas easily pass through. The ACNT
chip is simply put at the center of the plasma chamber. Other parameters used for
FIGURE 8.16 Positive SIMS spectrum of uncoated MWCNT.
10
4
2.0
1.0
0.0
10
4
1.5
1.0
0.5
0.0
10
2
4.0
2.0
0.0
10
2
0.8
0.4
0.0
10
1
4.0
2.0
0.0
10 20 30 40
50 60 70 80
90 100 110 120
130 140 150 160
170 180 190 200
C
1
C
4
C
6
C
7
H
7
+
C
10
H
8
+
C
6
H
4
-C
2
H
3
O
2
+
C
6
H
4
-C
4
H
5
O
2
+
C
5
C
2
C
3
I
n
t
e
n
s
i
t
y
Mass
ACNT processing are the same as for nanofiber processing, which were mentioned in
Section 8.4.
Figure 8.18(a) is the SEM image of the coated ACNTs. Compared to uncoated
ACNTs (Fig. 8.18(b)), one can see that the surface morphology exhibits smooth and
round curvatures, indicating coated polymer films.
HRTEM was used to further identify the coating layer, Figure 8.19(a) is the
HRTEM image of the uncoated ACNTs. In this figure, the silicon substrate, aligned
CNTs, and the nickel catalyst used in the CNT growth can be well identified. Figure
8.19(b) shows the coating layer clearly on the surfaces of the aligned CNTs. The
coating thickness is estimated to be about 1030 nm.
FIGURE 8.17 Positive SIMS spectrum of coated MWCNT.
10
4
3.0
2.0
1.0
0.0
10
4
10
3
2.0
1.0
0.0
3.0
2.0
0.0
10
2
6.0
4.0
2.0
0.0
10
2
8.0
6.0
4.0
2.0
0.0
10 0 20 30 40
50 60 70 80
90 80 100 110 120
130 120 140 150 160
160 170 180 190 200
I
n
t
e
n
s
i
t
y
Mass
40
H
+
C
+
CH
3
+
C
2
H
5
+
CF
+
C
2
F
+
CF
3
+
C
3
F
3
+
C
3
F
5
+
C
4
F
6
+
C
3
F
7
+
C
4
F
7
+
C
5
F
7
+
C
2
F
4
+
C
2
F
5
+
8.4 PROCESSING AND CHARACTERIZATION
OF NANOCOMPOSITE MATERIALS
Plasma coating of carbon nanofibers for enhanced dispersion and interfacial bonding
in polymer composites is discussed in this section. Carbon nanotubes can potentially
be used in many applications because of their desirable bulk properties.
1419
Recently
it has been shown in laboratory scale tests that the physical properties and perfor-
mance of polymer materials can be significantly improved by the addition of small
percentages of carbon nanotubes and nanofibers.
2023
However, there have not been
many successful tests that have used larger percentages of nanofibers as fillers. This
problem is associated with dispersing the nanofibers and creating a strong interface
between the nanofiber and the polymer matrix.
2425
The strong interface between
the nanofiber and the polymer matrix is essential to transfer the load from the matrix
to the nanofibers and thereby to enhance the mechanical properties of the composite.
In addition, the as-produced nanofibers usually form as aggregates that behave
differently in response to a load compared to individual nanofibers.
26,27
To maximize
the advantage of nanofibers as reinforcing particles in high-strength composites,
the aggregates need to be broken up and dispersed to prevent slippage.
FIGURE 8.18 HRTEM images of aligned carbon nanotubes: (a) coated with polystyrene;
(b) uncoated.
FIGURE 8.19 TEM images of (a) uncoated ACNT and (b) coated ACNT.
A key aspect of being able to manipulate the properties of the nanofibers is the
surface treatment of the nanofibers using various processing techniques. In the
previous section, we demonstrated the successful plasma deposition of a thin polymer
film on the surfaces of carbon nanofibers. In this section, we present the results on
the microstructure, dispersion, and mechanical properties of a polymer composite
impregnated with coated carbon nanofibers. The fracture morphology of both coated
and uncoated nanofiber composites have been identified by scanning electron
microscopy (SEM) observation using a Philips XL30 FEG SEM. High-resolution
transmission electron microscopy (HRTEM) images were acquired using a JEOL
2010F TEM to show the interface structures that are responsible for the improved
properties. TEM samples of the composite were prepared by ultramicrotomy with
a cutting thickness of 60 nm.
In this experiment, commercial Pyrograf III carbon nanofibers were used as
substrates. The Pyrograf III nanofibers are 70200 nm in diameter and 50100 Pm
long. Polystyrene is used as the monomer for the plasma polymerization because of
the good compatibility with the matrix polystyrene. The coating processing used
here was the same as the method mentioned in Section 4.1.3.4.
The procedure to form the nanocomposite is described next. Two grams of poly-
styrene powder was weighed and then mixed mechanically with coated or uncoated
nanofibers in appropriate proportions, i.e., 0 wt%, 1 wt%, 3 wt%, and 5 wt%. A solvent
(50 ml toluene) was then added to the premixed powders and the powder was thoroughly
dispersed ultrasonically using a low power bathtype sonicator. The solution was evap-
orated until its volume reduced to ~20 ml (the ultrasonic vibration was kept on during
this process) and then poured into an 80-mm u 60-mm u 6.5-mm aluminum mold. The
mold and the solution were kept at room temperature and dried for 7 days. After the
sample was completely dried, it was sectioned into 50-mm u 6-mm u 0.4-mm samples
for tensile testing according to ASTM D822-97, Standard Test Method for Tensile
Properties of Thin Plastic Sheeting.
An Instron mechanical testing machine, Model 2525-818, with a 1 mm/min
cross head speed was used for the tensile test.
28
Before the mechanical test, the first observation took place in the solution stage.
After ultrasonic mixing, the solutions with coated and uncoated nanofibers were
stored in small jars, as shown in Fig. 8.20. A stunning difference at this stage was
FIGURE 8.20 Solutions with coated (left) and uncoated (right) nanofibers.
that, as shown in this figure, the solution of uncoated nanofibers quickly separated
within 24 h (Fig. 8.20, right, nanofibers are precipitated from the solvent and
concentrated at the bottom of the jar), while the solution with coated nanofiber
remained homogenously suspended as shown in Fig. 8.20 (left) for as long as several
months. This phenomenon clearly indicates significantly modified surface behaviors
due to plasma coating.
Figure 8.21 shows the strength as a function of nanofiber concentration for both
coated and uncoated nanofiber composites. For the uncoated nanofiber composite,
the strength of the composite decreases gradually as the nanofiber concentration
FIGURE 8.21 Properties of a polystyrene nanocomposite: (a) the strength of composite VS
CNT concentration; (b) the modulus of composite VS CNT concentration.
Coated
Uncoated
22
20
18
16
14
12
10
8
0 1 2 3 4 5 6
Concentration (wt. %)
S
t
r
e
n
g
t
h

(
M
p
a
)
(a)
Coated
Uncoated
2.6
2.4
2.2
2
1.8
1.6
1.4
0 1 2 3 4 5 6
Concentration (wt. %)
M
o
d
u
l
u
s

(
G
p
a
)
(b)
increases, while the coated counterpart showed a significant increase in strength.
The maximum strength of the coated nanofiber composite takes place at 3 wt% and
then decreases up to 5 wt%. The modulus value is shown for both composites in
Fig. 8.19(b). A similar trend is seen, which is consistent with the strength values
(Fig. 8.21(a)). The decrease in properties above 3 wt% loading may be due to the
nanofibers not being initially well dispersed, and due to later agglomeration of the
nanofibers in the matrix. It is anticipated that the composite properties will mono-
tonically increase with the wt% loading of nanofibers if the dispersion can be
improved and maintained.
Figure 8.22 shows the fracture surfaces of the 3 wt% uncoated sample. The
nanofibers are highly clustered in the matrix with approximately a a10-Pm diameter
(Fig. 8.22(a)), as indicated by the arrows. These clusters appear to be densely distrib-
uted with a small spacing of a25 Pm (Fig. 8.22(b)).
Another important characteristic of the uncoated nanofibers composite is the
rather flat fracture surface (Fig. 8.22(c)), indicating the nature of brittle fracture. At
these fracture surfaces, severe pullouts of nanofibers are also observed, as shown in
Fig. 8.22(d). In sharp contrast, the dispersion is greatly improved in the coated
nanofibers composite. Figure 8.23 shows the fracture surfaces of the 3 wt% coated
nanofiber composite. The coated nanofibers are well dispersed (Fig. 8.23(a)) in the
matrix with a wavy type of fracture surface morphology (Fig. 8.23(b)). The interface
structure between the CNTs and polymer matrix was studied by HRTEM for both
coated (Fig. 8.24(a)) and uncoated nanofiber (Fig. 8.24(b)) composite samples. The
contrast in Fig. 8.24(a) clearly shows the coating layer between the carbon nanofiber
and the matrix, whereas the uncoated carbon nanofiber surface is in direct contact
with the matrix, as shown in Fig. 8.24(b).
FIGURE 8.22 SEM images of the fracture surfaces of the 3 wt% uncoated sample at different
resolution (a), (b), (c), and (d).
The central focus of this study was the enhanced interfacial bonding due to
plasma-coated thin films on nanofibers. The nature of strengthening in nanofiber-
reinforced composites is dependent on the stress transfer between the matrix and
nanofiber. For polymers, tensile loading can produce matrix cracking, nanofiber
bridging, nanofiber rupture, nanofiber pullout, and debonding. In this experiment,
pullouts of nanofibers were observed in the uncoated nanofiber composite, as indi-
cated in Fig. 8.22(d), especially within the cluster regions. As the nanofibers are
clustered, the interface area between the matrix and nanofiber is greatly reduced,
leading to significantly lowered strength. Furthermore, these clusters act as large
voids that are responsible for decreasing the strength of the composite as the nanofi-
ber concentration increases (Fig. 8.21).
As the nanofiber surfaces are modified by plasma coating, the surface energy
can be significantly lowered, which can enhance dispersion in the polymer matrix.
The well-dispersed nanofibers in the matrix appear to have few clusters and pullouts.
In addition, the adhesive film on the nanofiber surface, as shown in Fig. 8.24(a), can
FIGURE 8.23 SEM images of the fracture surfaces of the 3 wt% coated sample (a) and (b).
FIGURE 8.24 HRTEM image of the interface structure between the CNTs and polymer
matrix: (a) coated; (b) uncoated.
provide enhanced bonding, and therefore contribute to a considerably increased
strength in the coated-nanofiber composite. The efficiency of stress transfer is
strongly dependent on the maximum value of the shear stress acting at the interface.
This stress is also characterized as the interfacial shear strength that depends on the
nature of bonding at the interface. As indicated by the interface HRTEM, there is
clearly an interfacial adhesion layer due to the coated polymer film on the nanofiber
surface. Although a quantitative measure of the interfacial shear strength has not
been conducted, the effect of enhanced bonding is evident from the increased
composite strength and fracture surface morphology.
Although polystyrene is brittle, a number of studies showed that polystyrene
can be macroscopically toughened and manipulated to deform via shear yielding by
controlling the microstructure. The improvement of toughness of heterogeneous
polystyrene systems is mainly achieved by enhancing the strength of craze and thus
the craze resistance, or decreasing the concentrated stress. In the coated nanofiber
composite, the improved interfacial bond strength between the nanofiber and matrix
due to an adhesive thin film on nanofiber surfaces could increase the strength of the
craze. Furthermore, the increase of the interfacial adhesion may suppress the pro-
duction of voids or flaws in the polymer matrix, which might grow into cracks.
Thus, the fracture surface of the coated nanofiber composite exhibits typical shear
yielding behavior. In contrast, in the uncoated nanofiber-polymer composite, a rather
flat, brittle type of fracture surface occurs (Fig. 8.22(c)), similar to the fracture surface
of pure polystyrene. This behavior suggests that the highly clustered nanofibers in
the matrix do not contribute to shear yielding, and the uncoated nanofiber composite
shows a brittle fracture feature. In addition, the uncoated nanofibers may cause the
formation of voids (due to nanofiber pullouts) and defects. This could lower the stress
required for craze initiation and thus decrease the craze resistance, consistent with
the observation that the strength of the uncoated nanofiber-composite decreases with
the increasing percentage of nanofibers (Fig. 8.21).
Overall a unique approach has been developed to enhance the dispersion and
interfacial bonding of nanofibers in polymer composites. As a result of plasma
coating, carbon nanofibers can be well dispersed in a polymer matrix. Both the
fracture behavior and tensile strength data indicate that the well-dispersed nanofibers
have contributed to enhanced interfacial shear strength, and therefore have increased
the overall strength of the material. It is believed that the strength of the nanofiber
composite will also be enhanced based on the identification of the bonding mecha-
nisms between the nanofiber and coating, and the coating and the matrix material.
8.5 SUMMARY
Polymer thin films (such as acrylic acid, MMA, C
6
F
14
, and polystyrene) were depos-
ited on the surfaces of nanoparticles and nanofibers by a plasma polymerization
treatment. The average size of the nanoparticles was from several nanometers to
several hundreds of nanometers in irregular shapes. HRTEM experiments showed
that extremely thin films of the polymer (315 nm) were uniformly deposited on the
surfaces of the nanoparticles and nanofibers. The HRTEM results were confirmed by
time-of-flight secondary ion mass spectroscopy and IR. The deposition mechanisms
and the effects of plasma treatment parameters were discussed. The experiments
showed that the plasma processing is a powerful tool for surface modification of
nanoparticles. However, there are still several challenges that must be faced before
the plasma polymerization method can be widely used as a standard procedure for
nanoparticle surface modification.
The first problem is to design a more efficient stirring method to improve the
reproduction of the coating for each run. In plasma processing, if the substrate is a
flat surface, controlling the plasma processing parameters will give good control of
the film properties, including the thickness and the composition. This technique has
been widely used in the semiconductor industry for the surface modification of
wafers. But, for nanoparticles, the stirring method plays a very important role in the
overall processing. The methods used here are magnetic stirring and mechanical
stirring. The stirring parameters are difficult to control, which affects the control of
the coatings. One approach that is being investigated is to disperse the nanoparticles
in the monomer that is injected into the plasma system. In the future, an efficient
method to stir the powder will be a key for improvement of the plasma technique
when used for nanoparticle surface modification.
Another problem is standardization of the plasma system. The parameters for plasma
processing are highly system dependent. The parameters obtained from one system may
not be optimal when used in another system, even when the monomers and substrates
are exactly the same. This is a disadvantage for data sharing and comparison.
Finally, to take full advantage of the plasma surface coating technique, it is sug-
gested that many different monomers be investigated for use as coatings. With different
monomers, different experimental conditions also need to be tried in order to get a
uniform film on the nanosurfaces. Although polymers have been used widely in plasma
treatment, metals or conducting materials can also be used as coatings to explore how
the electrical conductivity and electrical properties of the coated nanoparticles change.
PROBLEMS
1. Briefly explain what plasma polymerization processing is and what the
important parameters for this processing are.
2. Compared with other functionalization methods for carbon nanotubes,
what are the advantages of using plasma processing?
3. A company has some amorphous carbon substrates. In order to modify
the surface, plasma polymerization processing is suggested. In order to
get better adhesion between the coating and substrates, which condition
will you will try to use for this processing: (a) high power, low flow ratio
or (b) low power, high flow ratio? If the substrate is polystyrene, which
condition you will use for this case? (Assume that the monomer used here
is styrene, and that other processing parameters are the same.)
4. A company wants to coat a glass substrate with poly-pyrrole to make the
surface ionically conductive. If the regular plasma polymerization pro-
cessing is used, could we achieve this objective? How could you modify
the plasma processing to achieve the objective?
5. What are the advantages and limitations of plasma polymerization pro-
cessing?
6. List as many parameters as you can think of that control the plasma
processing. Explain what effect you think each parameter would have on
the coating of nanoparticles. For example, faster or more vigorous stirring
of the nanoparticles may expose the particle to the plasma longer and make
the coating thicker and more uniform and prevent clumping of particles.
REFERENCES
1. Siegel, R.W., Nanostructured Materials-mind Over Matter, Nanostruct. Mater., 3(1),
118, 1993.
2. Hadjipanayis, G.C. and Siegel, R.W., Nanophase Materials, Synthesis-Properties
Applications, Kluwer Academic, Dordrecht, the Netherlands, 1994.
3. Whitesides, G.M., Mathias, J.P., and Seto, C.T., Molecular Self-Assembly and
Nanochemistrya Chemical Strategy for the Synthesis of Nanostructures, Science,
254, 13121319, 1991.
4. Stucky, G.D. and MacDougall, J.E., Quantum Confinement and Host/Guest Chem-
istry: Probing A New Dimension, Science, 247, 669671, 1990.
5. Gleiter, H., Nanostructured Materials: State of the Art and Perspectives, Nanostruct.
Mater., 6(1), 314, 1995.
6. Wolde, A.T., Ed., Nanotechnology, STT Netherlands Study Center for Technology
Trends, The Hague, the Netherlands, 1998.
7. Inagaki, N., Tasaka, S., and Ishii, K., Surface Modification of Polyethylene and
Magnetite Powders by Combination of Fluidization and Plasma Polymerization, J.
Appl. Polym. Sci., 48(8), 14331440, 1993.
8. Bayer, C., Karches, M., Mattews, A., and Von Rohr, P.R., Plasma Enhanced Chemical
Vapor Deposition on Powders in a Low Temperature Plasma Fluidized Bed, Chem.
Eng. Technol., 21(5), 427430, 1998.
9. van Ooij, W.J., Luo, S., Zhang, N., and Chityala, A., in Proceedings International
Conference on Advanced Mfg. Technology, Science Press, New York, 1999, p. 1572.
10. van Ooij, W.J. and Chityala, A., Surface Modification of Powders by Plasma Poly-
merization, Mittal, K.L., Ed., VSP, Utrecht, the Netherlands, 2000, pp. 243253.
11. van Ooij, W.J., Zhang, N., and Guo, S., in Fundamental and Applied Aspects of
Chemically Modified Surfaces, Blitz, J.P and Little, C.B., Ed., Royal Society of
Chemistry, Cambridge, U.K. 1999, pp. 191211.
12. Shi, D., He, P., Lian, J., Wang, L.M., and van Ooij, W.J., Plasma Deposition and
Characterization of Acrylic Acid Thin Film on Zno Nanoparticles, J. Mater. Res.,
17(10), 25552560, 2002.
13. Shi, D., Lian, J., He, P., Wang, L.M., van Ooij, W.J., Schulz, M., Liu, Y.J., and Mast,
D.B., Plasma Deposition of Ultrathin Polymer Films on Carbon Nanotubes, Appl.
Phys. Lett., 81(27), 52165218, 2002.
14. Iijima, S., Helical Microtubules of Graphitic Carbon, Nature, 354, 5658, 1991.
15. Baughman, R.H., Cui, C., Zakhidov, A.A., Iqbal, Z., Barisci, J.N., Spinks, G.M.,
Wallace, G.G., Mazzoldi, A., De Rossi, D., Rinzler, A.G., Jaschinski, O., Roth, S.,
and Kertesz, M., Carbon Nanotube Actuators, Science, 284, 13401344, 1999.
16. Gao, M., Dai, L., Baughman, R.H., Spinks, G.M., and Wallace, G.G., Electroactive
Polymer Actuators and Devices, SPIE Proceedings, 2000, p. 1824.
17. Hadjiev, V.G., Iliev, M.N., Arepalli, S., Nikolaev, P., and Files, B.S., Raman Scattering
Test of Single-Wall Carbon Nanotube Composites, Appl. Phys. Lett., 78(21),
31933195, 2001.
18. Liu, C., Cheng, H.M., Cong, H.T., Li, F., Su, G., Zhou, B.L., and Dresselhaus, M.S.,
Synthesis of Macroscopically Long Ropes of Well-Aligned Single-Walled Carbon
Nanotubes, Adv. Mater., 12(16), 11901192, 2000.
19. Walters, D.A., Casavant, M.J., Qin, X.C., Huffman, C.B., Boul, P.J., Ericson, L.M.,
Haroz, E.H., OConnell, M.J., Smith, K., Colbert, D.T., and Smalley, R.E., In-plane-
aligned Membranes of Carbon Nanotubes, Chem. Phys. Lett., 338(1), 1420, 2001.
20. Zhu, J., Kim, J.D., Peng, H.Q., Margrave, J.L., Khabashesku, V.N., and Barrera, E.V.,
Improving the Dispersion and Integration of Single-Walled Carbon Nanotubes in
Epoxy Composites through Functionalization, Nano Lett., 3(8), 11071113, 2003.
21. Tostenson, E.T. and Chou, T.W., Aligned Multi-Walled Carbon Nanotube-Reinforced
Composites: Processing and Mechanical Characterization, J. Phys. D, 35, L-77, 2002.
22. Cadek, M., Coleman, J.N., Barron, V., Hedicke, K., and Blau, W.J., Morphological
and Mechanical Properties of Carbon-Nanotube-Reinforced Semicrystalline and
Amorphous Polymer Composites, Appl. Phys. Lett., 81(27), 51235125, 2002.
23. Franklanda, S.J.V., Harikb, V.M., Odegarda, G.M., Brennerc, D.W., and Gatesd, T.S.,
The StressStrain Behavior of PolymerNanotube Composites from Molecular
Dynamics Simulation, Composites Science and Technology, 63(11), 16551661,
2003.
24. Lau, K.-T. Chem. Phys. Lett., 370, 399, 2003.
25. Mamedov, A.A., Kotov, N.A., Prato, M., Guldi, D.M., Wicksted, J.P., and Hirsch, A.,
Molecular Design of Strong Single-Wall Carbon Nanotube/Polyelectrolyte Multilayer
Composites, Nat. Mater., 1, 190194, 2002.
26. Ajayan, P.M., Nanotubes from Carbon, Chem. Rev., 99(7), 17871800, 1999.
27. Ajayan, P.M. and Zhou, O.Z., in Carbon Nanotubes, Dresselhaus, M.S., and Avouris,
Ph., Eds., Springer-Verlag, Berlin, 2001.
28. Shi, D., Lian, J., He, P., Wang, L.M., van Ooij, W.J., Schulz, M., Liu, Y.J., and Mast,
D.B., Plasma Coating of Carbon Nanofibers for Enhanced Dispersion and Interfacial
Bonding in Polymer Composites, Appl. Phys. Lett., 83(25), 53015303, 2003.
9
Structural
Nanocomposites
Hassan Mahfuz
CONTENTS
9.1 Introduction
9.2 Matrix Modification
9.2.1 Ultrasonic Mixing
9.2.2 Manufacturing of Nanocomposites
9.2.3 Micron- and Nanosized SiC Particle Infusion
9.2.4 Thermal Analysis of Carbon/ Epoxy Nanocomposites
9.2.5 Mechanical Tests
9.2.6 Fatigue Tests
9.3 Nanophased Filaments
9.4 Core Modification
9.5 Summary
Problems
References
9.1 INTRODUCTION
The subject of nanocomposites is especially interesting because at least one of a
nanocomposites phases has one or more dimensionslength, width, or thickness
in the nanometer range, which is usually defined as 1 to 100 nm. This is the range
where phenomena associated with atomic and molecular interactions strongly influ-
ence the macroscopic properties of the materials, but this is also the length scale
where our knowledge of how to synthesize and process materials is the weakest.
Nevertheless, it is very well known that the catalytic, mechanical, electronic, optical,
and other properties of a material can significantly and favorably be altered when
that material is fashioned from nanoscale building blocks.
16
For instance, nano-
crystalline copper is up to five times harder than conventional copper, and ceramics,
which normally are brittle, can be made more easily deformable if their grain size
is reduced to low nanometer range. Such improved properties can also be incorpo-
rated into nanocomposites, in which the building blockssay, nanoscale metal or
carbon particles, or nanometer-thick filaments or sheets of a ceramicare dispersed
in a matrix of another material, such as a polymer. Because the building blocks of
a nanocomposite are nanoscale, they have an enormous surface area, and therefore
significantly create large interfaces between the two intermixed phases. The special
properties of the nanocomposite arise from the interaction of its phases at these
interfaces. The challenge, however, lies in fabricating these materials with the
required atomic specifications.
One of the important areas for future investments in research identified by the
National Nanotechnology Initiative (NNI) is called beyond nano, which notes that
the advances at the nanoscale will be meaningless if they cannot be interfaced well
with the technology at larger material components, systems, and architectures to
produce usable devices.
7
One of the primary objectives of the NNI is to integrate
nano-objects and nanoscale phenomena into larger hierarchical systems. Develop-
ment of large structural-level laminates from nanoinfused polymers is a step in the
right direction.
With our continuing quest for lighter and stronger composites, the demand for
new types of materials is increasing. No longer can traditional fibrous composites
fulfill our stringent requirements, nor can they be engineered at the continuum level,
which controls properties at the molecular or atomic level. However, it is well known
that molecular forces and bonding, the interaction between the interfaces, and the
physical phenomena occurring at this level will dictate the aggregate properties of
materials. In this nanocrystalline state, the solids contain such a high density of
defects that the spacing between them approaches interatomic distances and a large
fraction of the atoms sits very close to a defect.
810
Consequently nanocrystalline
materials are exceptionally strong, hard, and ductile at high temperatures, wear
resistant, corrosion resistant, and chemically very active.
11,12
It has been established in recent years that polymer-based composites reinforced
with a small percentage of strong fillers can significantly improve the mechanical,
thermal, and barrier properties of pure polymer matrix.
1318
Moreover, these improve-
ments are achieved through conventional processing techniques without any detrimental
effects on processability, appearance, density, or aging performance of the matrix. These
composites are now being considered for a wide range of applications including the
packaging, coating, electronics, automotive, and aerospace industries. While nano-
particles have attractive attributes, their use in structural composites, which are
relatively large in dimension, is almost nonexistent.
In recent years, studies have been conducted to see how traditional and nontra-
ditional modifiers influence the thermal and mechanical properties.
19
A conventional
modifier, microscale nylon particles, and nonconventional modifiers, nanoclays with
different surface chemistries, were incorporated into an epoxy matrix used for
prepeg-based, fiber-reinforced composites. The incorporation of particle modifiers
resulted in changes in the interlaminar shear stress (ILSS) and char yield. In various
other studies,
2025
it has been established that the addition of a small amount of
nanoparticles (5 wt%) to a matrix can increase the strength, toughness, dimensional
stability, and resistance to thermal degradation without compromising the weight or
processability of the composite. This concept of infusion of nanoparticles into a
polymer by low volume or weight fraction forms the foundation of the manufacturing
of structural nanocomposites. Three routes are generally followed: (1) modification
of the matrix through particle infusion and reinforcement with regular fibers;
(2) dispersion of acicular nanoparticles such as carbon whiskers, carbon nanotubes
(CNT), or carbon nanofibers (CNF) into a textile polymer precursor, and melt extru-
sion into filaments; and (3) infusion of nanoparticles into liquid polymer foam and
construction of sandwich panels using nanophased foam materials. A detailed
description of the first route and a brief description of the other two routes are given
in the following sections.
9.2 MATRIX MODIFICATION
A systematic study has been carried out to investigate matrix properties by intro-
ducing micro- and nanosized SiC fillers into an epoxy matrix. The study has revealed
that with equal amount of loading, nanoparticle infusion brings about better thermal
and mechanical properties to the matrix than what is usually given by the microfill-
ers infusion. The nanophased matrix is then utilized in a vacuum-assisted resin
transfer molding (VARTM) set up with satin weave carbon fiber preforms to fabricate
laminated composites. The resulting structural composites have been tested under
flexural and tensile loads to evaluate mechanical properties. The fillers were nano-
and micron-size silicon carbide particles that were mixed with the SC-15 epoxy
resin using an ultrasonic processor. The amount of particle loading varied from 1.5
to 3.0% by weight of the resin. Ultrasonic mixing utilized high-energy sonic waves
to force an intrinsic mixing of particles with the matrix via sonic cavitations. In
parallel, control panels were also fabricated without particle infusion. It has been
observed that nanoparticle infusion increases the thermal stability of the system by
enhancing cross-linking in the polymer. Nanoparticles also tend to reduce void content
of the as-fabricated composites and thus translate into increased mechanical properties.
With 1.5 wt% loading, an average of 2030% increase in mechanical properties has
been observed. Fatigue tests were also performed under flexural loading, and the
performance of the nanoinfused system was seen to be superior to that of the neat
system.
9.2.1 ULTRASONIC MIXING
Various techniques such as melt mixing, solution mixing, shear mixing, and mechan-
ical stirring are employed to infuse nanoparticles into a polymer. In addition to these
processes, acoustic cavitation is also one of the efficient ways to disperse nanopar-
ticles into virgin materials.
26
In this case, the application of alternating acoustic
pressure above the cavitation threshold creates numerous cavities in the liquid. Some
of these cavities oscillate at a frequency of the applied field (usually 20 kHz), while
the gas content inside these cavities remains constant. However, some other cavities
grow intensely under tensile stresses, while yet another portion of these cavities,
which are not completely filled with gas, start to collapse under the compressive
stresses of the sound wave. In the latter case, the collapsing cavity generates tiny
particles of debris and the energy of the collapsed one is transformed into pressure
pulses. It is noteworthy that the formation of the debris further facilitates the devel-
opment of cavitation. It is assumed that acoustic cavitation in liquid develops accord-
ing to a chain reaction. Therefore, individual cavities on real nuclei develop so
rapidly that within a few microseconds an active cavitation region is created close
to the source of the ultrasound probe. The development of cavitation processes in
the ultrasonically processed melt creates favorable conditions for the intensification
of various physiochemical processes. Acoustic cavitation accelerates heat and mass
transfer processes such as diffusion, wetting, dissolution, dispersion, and emulsifi-
cation. Recently it has been reported that there is a clear acceleration of polymer
reaction under ultrasound in both catalysed and uncatalysed reactions.
27
In the
present investigation, ultrasonic mixing was employed to infuse SiC nanoparticles
into part A of the SC-15 epoxy resin.
9.2.2 MANUFACTURING OF NANOCOMPOSITES
The fabrication of nanophased carbon/epoxy composites was carried out in three
steps. In the first step, spherical SiC nanoparticles of about 29 nm in diameter
(manufacturer: MTI Corporation, Inc., Richmond, CA) were ultrasonically mixed
with part A (mixture of diglycidylether of bisphenl A, 60 to 70%, aliphatic digly-
cidylether, 10 to 20%, and epoxy toughner, 10 to 20%) of SC-15 epoxy resin
(manufacturer: Applied Poleramic, Inc., Benicia, CA). SC-15 is a two-phase tough-
ened epoxy resin system that cures at room temperature and is extensively used in
vacuum-assisted resin transfer molding (VARTM) processes. The loading of nano-
particles ranged from 1.5 to 3.0% by weight of the resin. The mixing was carried
out in a sonics vibra cell ultrasonic liquid processor (Ti-horn, frequency 20 kHz,
intensity 100 W/cm
2
), as shown in Fig. 9.1. The mixing was carried out at 55%
of the amplitude for about 30 min. At this time, the dispersion of nanoparticles
seemed uniform through visual observation. In order to avoid rise in temperature
during sonication, cooling was employed by submerging the mixing beaker in a
liquid chamber where temperature was autocontrolled through a computer. In the
next step, Part B (hardener, cycloaliphatic amine, 70 to 90% and polyoxylalkylamine,
10 to 30%) was added with the mixture at a ratio of 3: 10 and the mixing was carried
FIGURE 9.1 Vibra-cell ultrasonic processing.
out mechanically for about 10 min using a high-speed mechanical stirrer. In the final
step, the reaction mixture was used in a VARTM set up
2830
with satin weave carbon
fiber preforms to fabricate carbon/epoxy nanocomposite panels, as shown in Fig.
9.2. In order to carry out studies with particle size, micron-sized SiC particles
(average particle size ~1 Pm; supplier: Atlantic Equipment Engineers, Bergenfield,
NJ) were also infused with SC-15 parallel to nanoparticle infusion. Modified part
A was then mixed with part B, and was cast into rectangular molds to cure to solid
panels. Test coupons were extracted from each category of panels to conduct various
chemical and mechanical tests.
9.2.3 MICRON- AND NANOSIZED SiC PARTICLE INFUSION
As stated earlier, the investigation began with the infusion of nano- and micron-
sized particles into SC-15 epoxy matrix. Tensile response of these systems is shown
in Fig. 9.3. As the figure shows, epoxy with nanoinfusion has the highest strength
and stiffness when compared with the other three systems, namely neat epoxy, epoxy
irradiated with ultrasonic cavitation, and epoxy with 1.5 wt% of micron-sized SiC
particles. As stated earlier, a batch of epoxy matrix was also irradiated for equal
length of time without any particle infusion to evaluate the effect of sonication on
the epoxy. Figure 9.3 and Table 9.1 show that sonicated epoxy performs as well as
the micron-SiC infused system, and much better than the neat epoxy. We see that
sonication alone has contributed to about 18% and 3% increase in stiffness and
strength, respectively. The reason for such change in stiffness of the matrix due to
ultrasound irradiation comes from the fact that the ultrasound irradiation enhances
the homogeneity of the reaction mixture part A of the SC-15 epoxy resin. As we
mentioned earlier, part A of SC-15 epoxy resin contains the mixture of diglycidyl
ether of bisphenol A, aliphatic diglycidyl ether, and epoxy toughner. The ultrasound
irradiation helps in molecular mixing of these components and the formation of
FIGURE 9.2 A typical VARTM setup.
FIGURE 9.3 Tensile response of SiC infusion in SC-15 epoxy.
TABLE 9.1
Tensile Response of Neat and Nanoinfused SC-15 Epoxy Composites
Material
Tensile Modulus
(GPa)
Gain/Loss
Modulus
(%)
Tensile Strength
(MPa)
Gain/Loss
Strength
(%)
Neat Epoxy 2.16 2.14 57.30 55.00
2.12 55.20
2.13 52.50
Sonicated Epoxy 2.54 2.53 18.22 57.13 56.57 2.85
2.50 55.46
2.55 57.14
1.5% Pm SiC-Epoxy 2.60 2.55 19.15 56.52 56.21 2.20
2.5 55.91
2.54 56.21
1.5% Nano SiC-Epoxy 3.10 3.10 44.85 60.87 63.70 15.81
3.21 62.54
3.00 67.68
5
Strain (%)
0
20000
40000
60000
80000
S
t
r
e
s
s

(
K
P
a
)
Plot legend
Neat
Sonicated epoxy
+1.5% SiC nano
+1.5% SiC micron
0 1 2 3 4
reactive species, which ultimately leads to an increase in the cross-linking in the
polymer when mixed with part B of SC-15. This effect is more prominent when
each of the highly reactive surfaces of the nanoparticle is efficiently coated with part
A of SC-15 resin and reacted with the hardener, part B of SC-15 resin. When the
matrix is sonicated with SiC nanoparticles in it, it is observed that the enhancement
in stiffness and strength significantly increases to around 45% and 13%, respectively,
over the neat system. The corresponding improvement over the sonicated-epoxy
system, however, is 22% and 12%, respectively, which we believe is entirely due to
the nanoparticle infusion. On the other hand, the infusion of micron-sized particles
with identical loading (1.5 wt%) did not contribute to any improvement of properties
over the sonicated-epoxy system. Although we see some improvements in properties
with respect to the neat system, the contribution is actually coming from the sonic
cavitation effect rather than from the particle infusion. We also note from Fig. 9.3
that the percent strain to failure of the nanoinfused system reduces to 2.2 from around
3.0, which is typical with the other systems. A quick calculation based on 1.5 wt%
loading shows that the particle volume fraction in the composite in both cases is
around 0.53%. With such low particle volume fraction, it is very difficult to visualize
a 44% increase in stiffness of the nanoparticle-infused system. The underlying
phenomenon thus points to interatomic/molecular activities in the polymer that is
taking place in the presence of the nanosized particles. The small size of these
particles effectively increases the surface energy significantly, and subsequently
enhances crystallinity and cross-linking of the polymer. This in turn translates into
improved stiffness and strength of the resulting composites. The overall inference
from the study was that with low amounts of loading in an epoxy matrix, nanopar-
ticles offer significant benefits over the parallel microparticle infusion.
9.2.4 THERMAL ANALYSIS OF CARBON/EPOXY NANOCOMPOSITES
In the next phase of the investigation, epoxy matrix infused with 1.5 wt% of SiC
nanoparticles were reinforced with satin woven carbon fibers in a traditional com-
posite manufacturing procedure (VARTM). Coupons extracted from the resulting
fibrous composite laminates were then subjected to thermal and mechanical tests.
Thermogravimetric analysis (TGA) of various specimens was carried out under nitro-
gen gas atmosphere at a heating rate of 10qC/min on a TA Instruments, Inc. apparatus.
The real-time characteristic curves were generated by the Universal Analysis 200
(TA Instruments, Inc.) data acquisition system. Data from a typical thermogravimetric
analysis of the fibrous composite are shown in Fig. 9.4(a).
In an attempt to determine the effect of the amount of particle loading, a separate
batch of materials was manufactured with 3 wt% in addition to 1.5 wt% loading.
The percent weight loss with temperature of these two systems, along with the neat
composite, is shown in Fig. 9.4(a). In the present study, 50% of the total weight loss
is considered the structural destabilization point of the system. It is also common
practice to consider 50% weight loss an indicator for structural destabilization as
given in References 31 and 32. Figure 9.4(b) clearly shows that the neat sample
without SiC nanoparticles is stable up to 355qC, whereas in Fig. 9.4(c) shows the
samples with 1.5% loading are stable up to 365qC. The reason for this increase in
the thermal stability is due to the increase in cross-linking of the epoxy resin in the
presence of SiC nanoparticles and having minimum particle-to-particle interaction.
As the loading of SiC particles increases to 3%, as shown in Fig. 9.4(d), the thermal
stability drops down to 342qC, which is lower than that of the neat carbon epoxy
system shown in Fig. 9.4(b). The reason for the lowering of the temperature of the
3 wt% carbon/epoxy system may be explained macroscopically as a simple colliga-
tive thermodynamic effect of an impurity on a bulk solution. Microscopically it may
be seen as the result of the perturbation that the SiC introduces to the three-
dimensional structure of the polymer. This perturbation weakens the van der Waals
interaction between the polymer chains. This affects the stability of the polymer,
which is reflected in the lowering of the thermal stability. We believe this perturbation
FIGURE 9.4(a) TGA comparison graph of neat vs. nanophased systems.
FIGURE 9.4(b) TGA response of neat carbon/epoxy.
0 200 400 600 800
Temperature (C)
70
80
90
100
W
e
i
g
h
t

(
%
)
Plot legend
Neat
+1.5% SiC
+3.0% SiC
0
70
80
90
100
200 400 600 800
Temperature (C)
W
e
i
g
h
t

(
%
)
Carbon/epoxy
Neat
(354.52C, 87.07%)
87.07%
(1.91 mg, 27.42%)
311.14C
399.51C
begins at a point when the number of particle reaches a certain level and particle-
to-particle interaction initiates leading to agglomeration of particles into lumps,
which acts as an impurity in the system.
Thermal mechanical analysis (TMA) of the various systems is shown in Fig. 9.5.
Figure 9.5 shows that the dimensional changes are more or less identical with the
three systems until one reaches the Tg of the material. Beyond this temperature,
dimensional change with 1.5 wt% system is quite lower than the neat and 3 wt%
system. This dimensional change is also a measure of the coefficient of thermal
expansion (CTE) of the material system being tested. Since the dimensional change
was measured in the thickness direction of the laminate, the slope of the curves in
FIGURE 9.4(c) TGA response of carbon/ epoxy with 1.5% SiC nanofillers.
FIGURE 9.4(d) TGA response of carbon/epoxy with 3.0% SiC nanofillers.
0 200 400 600 800
Temperature (C)
70
80
90
100
W
e
i
g
h
t

(
%
)
Plot legend
+1.5% SiC
(364.79C, 87.04%)
(26.65%, 1.421 mg)
321.35C
410.58C
0 200 400 600 800
Temperature (C)
70
80
90
100
W
e
i
g
h
t

(
%
)
Plot legend
+3.0% SiC
(341.55C, 89.20%)
(25.61%, 1.334 mg)
299.24C
384.04C
Fig. 9.5 accordingly indicates CTE in that particular direction. As Fig. 9.5 shows, there
is a slight reduction in CTE of the 1.5 wt% system over the neat and 3 wt% systems.
9.2.5 MECHANICAL TESTS
Two types of mechanical tests, namely flexure and tensile, were performed to evaluate
the bulk stiffness and strength of each of the material systems. A typical stress strain
behavior from the flexural test is shown in Fig. 9.6. It is observed that the system with
FIGURE 9.5 TMA of carbon/ epoxy nanocomposites.
FIGURE 9.6 Flexural testing of carbon/epoxy nanocomposites.
z
z
z
z
z
z
z
z
z
0
50
50
100
150
200
D
i
m
e
n
s
i
o
n

c
h
a
n
g
e

(
m
)
50 0 50 100 150 200 250 300 350
Temperature (C)
z

Neat-carbon-SC15
1.5% SiC-carbon-SC15
3% SiC carbon-SC15
0 0.004 0.008 0.012 0.016
Strain (mm/mm)
0
200
400
600
S
t
r
e
s
s

(
M
P
a
)
Flexure test comparison
(plot legend)
Neat sample
1.5% SiC sample
3% SiC sample
1.5% infusion has the highest strength and stiffness among the three systems
indicated in Fig. 9.6. Gain in strength and stiffness of the 1.5% system is around
30% and 12%, respectively, over the neat, as shown in Table 9.2. Enhancement in
strength by 30% during flexure was somewhat surprising due to the fact that
previous studies with nanoclays
19,20
did not show such improvement. As Fig. 9.3
and Table 9.1 show, if the particle infusion is increased to 3%, there is no propor-
tional improvement in properties. Rather, there is a very nominal increase in strength
and a significant reduction in stiffness with the 3% wt system. Similar reflection
in properties with somewhat different ratios is observed during tensile tests, as
shown in Fig. 9.7 and Table 9.3. It is noted that the enhancements in strength and
stiffness are consistent with those shown in Fig. 9.3. Large values of strength and
modulus associated with Fig. 9.7 with respect to those in Fig. 9.3 are due to the
reinforcement of fibers. The fiber reinforcement, however, did not affect the strain
to failure of the 1.5% system, as it is seen to be around 2.25% (Fig. 9.7), which is
almost identical to that indicated in Fig. 9.3. Possible reasons for such behavior
may be that since the reinforcement is in cloth (satin woven) form, it did not
contribute much to the elongation, and that the failure was mostly controlled by
matrix- and delamination-related failure modes. Correspondingly, enhancement in
the case of fibrous composites (Fig. 9.7) is very similar to that obtained with
nanophased matrix (Fig. 9.3).
In an attempt to understand the gain in properties in both cases (flexure and
tension), TEM micrographs were taken and investigated for the nanophased systems.
TABLE 9.2
Flexural Test Data for Carbon/Epoxy Composites
Material
Flexural
Strength
(MPa)
Gain/Loss
Strength
(%)
Flexural
Modulus
(GPa)
Gain/Loss
Modulus
(%)
Neat 334.0 381.8 45.0 45.6
390.0 46.0
395.0 39.0
390.0 44.0
400.0 54.0
1.5% SiC 550.5 531.3 39.2 50.3 51.26 12.4
505.0 51.0
525.0 47.0
556.0 59.0
520.0 49.0
3.0% SiC 424.1 399.8 4.7 37.5 38.1 16.4
390.0 36.0
395.0 39.0
390.0 41.0
400.0 37.0
FIGURE 9.7 Tensile testing of carbon/epoxy nanocomposites.
TABLE 9.3
Tensile Test Data for Carbon/Epoxy Composites
Material
Tensile
Strength
(GPa)
Gain/Loss
Strength
(%)
Tensile
Modulus
(GPa)
Gain/Loss
Modulus
(%)
Neat 1.196 1.19 50.1 50.1
1.192 51.5
1.193 39.0
1.194 53.9
1.194 55.9
+1.5% SiC 1.330 1.33 +11.6 63.2 61.8 +23.5
1.430 59.5
1.320 61.9
1.290 62.8
1.290 61.8
+3.0% SiC 0.500 0.48 53.9 53.1 54.0 +6.3
0.500 51.3
0.450 61.0
0.380 50.4
0.550 54.0
0 0.005 0.01 0.015 0.02 0.025
Strain (mm/mm)
0
0.4
0.8
1.2
1.6
S
t
r
e
s
s

(
G
P
a
)
Plot legend
Carbon/epoxy composite
Neat
+1.5% SiC
+3.0% SiC
One such TEM of the fibrous composite with 1.5% infusion is shown in Fig. 9.8.
It can be seen in Fig. 9.8 that the fiber matrix interface is very uniform, strongly
bonded with the matrix, and is almost undisturbed by the infusion of nanoparticles.
Nanoparticles are also uniformly dispersed over the entire body of the matrix.
Micron-size voids are also seen in the figure, some of which are partially filled with
nanoparticles. An enlarged view of these voids and nanoparticles is shown in Fig. 9.9.
Void content tests performed on the nanophased composites revealed that there is a
reduction in void content by about 3040% due to 1.5% infusion. We believe this
reduction in the void content of the matrix provided strength for the nanophased
composites. When the infusion increased to 3%, as is observed in the TEM micro-
graph of Fig. 9.10, the particles started to form lumps that sometimes became larger
FIGURE 9.8 TEM micrograph of carbon/epoxy nanocomposites.
FIGURE 9.9 TEM Micrograph of nanoparticles and voids at 1.5% wt loading.
than the size of the voids. Instead of filling voids, these lumps acted almost like
impurities in the polymer. We believe with 3% loading, the number of nanoparticles
are very high, which allows particle-to-particle interaction rather than the intended
particle-to-polymer interaction. Once it reaches this state, the particles begin to
agglomerate and form lumps that eventually affect the van der Waals interaction
between the polymer chains, and reduce the cross-linking and increase void content
in the nanocomposite. The resulting mechanical properties are hence degraded. This
effect can be clearly visualized from careful observation of TGA data from 600qC
to 800qC in Fig 9.4. Figure 9.4(a) and (b) shows that the percentage weight losses
within the temperature range of 600qC to 800qC are about 1% and 0.5%, respectively,
for neat carbon epoxy and 1.5% SiC carbon epoxy samples, whereas Fig. 9.4(c)
indicates that the percentage weight loss at the same temperature range is about
2.5%. It is also seen in Fig. 9.4(c) that the curve continues to fall further beyond
800qC. The possible reasons for these relatively high percentage weight losses could
be due to the continuous loss of void content in the system. In other words, this
suggests the presence of higher void content in the 3% SiC carbon epoxy system
than the neat and 1.5 wt% SiC carbon epoxy systems.
9.2.6 FATIGUE TESTS
Since property enhancement was positive from a quasistatic point of view, it was
decided to test these materials under dynamic loading such as fatigue. Accordingly
flexural fatigue tests were conducted at a stress ratio of 0.1 and a frequency of 3 Hz.
S-N diagrams, shown in Fig. 9.11 were then generated for the three systems. We
observe in general that above 60% load level, the neat system performs better than
any of the nanoinfused systems. The 3% system demonstrates a very poor fatigue
performance over the entire range of loading, which is similar to the 3 wt% system
during quasistatic tension and flexure, as we saw earlier. During fatigue, once the load
FIGURE 9.10 Nanoparticle agglomeration at 3 wt% loading.
level goes below 60%, the fatigue performances of the neat and 1.5% system begin
to reverse. For example, at 65% load level, the ratio of cycle numbers to failure between
the neat and the 1.5% system is around 1.75, whereas this ratio changes to 0.56 at
55% load level. The reversal of the fatigue phenomenon is not yet fully understood.
However, as shown in the S-N diagram, the slope of the neat system is much steeper
that that of the 1.5% system, suggesting a lesser sensitivity of the neat system with
respect to applied load. This indicates that for a change in the load level, the corre-
sponding change in cycle numbers will be much smaller in the case of the neat system.
Since this continued up to 50% load level, we were able to find an intersection point
at around the 60% marker, which was defined as a threshold load level. Below this
threshold stress level, we believe the fatigue failure mechanisms such as matrix cracks,
filament splitting, and delamination
3335
are significantly slowed down with the 1.5%
system. And that is quite possible, since nanoinfusion (1.5%) certainly improves the
matrix properties and matrix-dominated failure modes, as we saw earlier. In this
investigation, we did not define any runout cycle numbers as is typical with fatigue
studies. Data shown in Table 9.4 represent all failed samples. Although it took a good
deal of time to conduct fatigue tests, especially at lower load levels, it was done
intentionally to capture the threshold stress level. It should also be noted in Fig. 9.11
that the slopes of the two nanophased systems (1.5% and 3%) are similar, indicating
that they are more or less equally sensitive to applied load with the exception that the
1.5 wt% system is outperforming the 3 wt% system by a large margin.
FIGURE 9.11 S-N diagram (flexural fatigue).
100 1000 10000 100000 1000000 10000000
Cycles
0.4
0.5
0.6
0.7
0.8
0.9
N
o
r
m
a
l
i
z
e
d

l
o
a
d

l
e
v
e
l
Carbon epoxy composite
Plot legend
Neat
+1.5% SiC
+3.0% SiC
9.3 NANOPHASED FILAMENTS
In another route, nanoparticles with aspect ratios such as carbon nanowhiskers and
carbon nanotubes were doped with textile polymer precursors. Two types of polymer
precursors were used: linear low-density polyethylene and nylon-6. The fillers in
each case were carbon nanowhiskers and multiwalled carbon nanotubes (MWCNT).
The amount of loading varied 12% by weight of the base polymer. Nanoparticles
were first mixed with the base polymer in powdered form. The mixture was then
dried in a hopper and fed through a single screw extrusion machine. It passed through
three stages of heating where the temperature was set slightly above the melting
temperature of the polymer. After heating and at the end of the screw, the molten
mass was forced through two stages of mixing. In the final stage, filaments were
extruded through a micron-size orifice. The extruded filaments went sequentially
through a cooling trough, a tension device, a heater, and eventually wound into a
filament winder to form into spools. In one attempt, these filaments were cut into
strands, laid out in 0q fashions in several layers, and consolidated in a compression
molding machine. Test coupons were extracted from these laminates and were tested
under tension and flexure. It was found that the improvement with carbon nano-
whiskers was around 17%, while it increased to 34% with MWCNT. Out of the four
systems (two fillers and two polymers) investigated, the system with nylon-6 infused
with MWCNT yielded the most promising results. Tension tests on individual fila-
ment of this system, shown in Fig. 9.12, demonstrate that there is about 150300%
TABLE 9.4
Flexural Fatigue Response (Frequency 3 Hz, R 0.1)
Type
Load Level
50% 57% 70% 80% 85% 90%
Neat 3,109,800 496,090 621,000 66,260 69,260 38,824
3,187,786 509,136 500,000 155,577 165,577 40,989
560,050 244,900 37,503 37,503 41,860
278,589 386,579 386,579
Average 3,148,793 521,759 411,122 161,480 164,730 40,558
1.5% SiC 5,090,088 2,945,712 200,000 416,285 1284 1604
5,146,479 120,281 39,350 55,327 1576 538
2,607,053 166,393 63,490 608 1066
105,525 52,744 3575
300,087 1400
Average 5,118,284 1,891,015 127,817 177,587 1689 1069
3.0% SiC 29,897 4688 200,000 643 286 121
27,765 5890 39,350 534 101 249
28,578 3420 166,393 765
29,657 105,525 1065
Average 28,974 4666 127,817 752 194 185
improvement in strength and stiffness with 1% MWCNT loading. TEM studies
revealed that extrusion technique caused sufficient alignment of MWCNT along the
length of the filament, which may have caused the gain in mechanical properties.
36
9.4 CORE MODIFICATION
In the third route, an innovative technique to develop polyurethane foams containing
nanoparticles was introduced. Polymethylene polyphenylisocyanate (part A) was
mixed with nanoparticles such as SiC and TiO
2
, and irradiated with a high-power
ultrasound liquid processor. In the next step, the modified foams containing nano-
particles were mixed with part B (containing polyol resin systems, surfactant, and
an amine catalyst) through a high-speed mechanical stirrer. The mixture was then
cast into rectangular molds to make nanophased foam panels. Test coupons were
then extracted from the panels to carry out thermal and mechanical characterizations.
The as-prepared foams were characterized by scanning electron microscopy (SEM),
x-ray diffraction, and thermogravimetric analysis (TGA). SEM studies showed that
the particles were nonagglomerated and well dispersed in the entire volume of the
foam. The foam cells were well ordered and uniform in size and shape. The TGA
analyses indicate that the modified foams were thermally more stable than the parallel
neat system.
37
Quasistatic flexure tests under three-point bend configuration were
also conducted with both modified and neat foams. Test results showed a significant
increase (approximately 5070%) in the flexural strength and stiffness of the
nanophased foams over the neat system. This enhancement in flexural properties was
demonstrated repeatedly with multiple batches and with at least three specimens
tested from each batch. The nanophased foam was then used with regular S-2 glass
FIGURE 9.12 Stress versus strain for neat Nylon 6 and Nylon 6 with 1 wt% multiwalled
carbon nanotubes with 102-mm gage length.
0
50
100
150
200
250
300
350
400
0.0 0.5 1.0 1.5 2.0 2.5 3.0
E
n
g
i
n
e
e
r
i
n
g

s
t
r
e
s
s

(
M
P
a
)
Engineering strain
Nylon 6 with 1 wt% multi-walled carbon nanotubes
Neat Nylon 6
fiber preforms and SC-15 epoxy to manufacture sandwich composites in a VARTM
setup. A parallel set of control panels was also made with neat polyurethane core
materials. A significant improvement in flexural strength and stiffness was observed
with 3% loading of TiO
2
nanoparticles in part A. Debond fracture toughness param-
eters (G
c
) were also determined for both categories of sandwich constructions, and
it was seen that nanoparticle infusion reduces the value of G
c
by almost a factor of
three. Despite this reduction, the strength of nanophased sandwich increased by
about 53% over the neat system, as shown in Fig. 9.13.
9.5 SUMMARY
With identical loading fraction, nanoparticle infusion into an epoxy system is seen
to be significantly superior to micron-sized particle infusion in enhancing the
thermal and mechanical properties of the resulting composites. A low-cost but
reliable manufacturing procedure can be used to fabricate large-scale laminated
structural composites with a nanophased matrix. Nanoparticles loading into the
base matrix are found to be optimal around 1.5% by weight to derive maximum
gain in the mechanical and thermal properties of the structural composites. Nano-
particle infusion is seen to increase the thermal stability of the matrix by enhancing
cross-linking in the polymer. The additional cross-linking seems to be due to the
catalytic effects caused by the presence of the nanoparticles. Nanoparticles tend to
act as fillers in the voids, which in turn reduces the void content of the composites
and thus translates into enhanced mechanical properties. With 1.5% loading in the
matrix, an average 2030% increase in mechanical properties of the resulting
fibrous composites has been observed both in tension and flexure. During flexural
fatigue, a threshold load level around 60% of the ultimate flexural strength seems
to exist. Below this threshold load level, the 1.5% system supersedes the neat
FIGURE 9.13 Flexural tests of neat and nanophased sandwich composites.
0 4 8 10 12 14 16
0
200
400
600
800
1000
Sandwich with nanophased core (TiO
2
)
Density 88 Kg/m
3
Sandwich with neat polycore
L
o
a
d

(
N
)
Displacement (mm)
2 6
system, whereas the situation is reversed if stress level is above this level. Through-
out the entire loading range, the 3% system is seen to be inferior to the other two
systems.
It is observed from our investigation that in textile precursors as well as in liquid
foam there are significant improvements in chemical and mechanical properties of
the resulting nanocomposites due to nanoparticle infusion. This improvement takes
place with a very low loading of nanoparticle, and there seems to be an optimal
loading corresponding to each system of nanoparticle and the polymer. But, in each
route and in each system, the optimal loading of nanoparticles of either spherical
or acicular shape is seen to be between 1 and 3% by weight. It is also noted that
the gain in mechanical properties is much more significant if one goes through the
fiber and the core modification routes.
PROBLEMS
1. Explain different methods that are commonly used in mixing nanosized
particles into polymers.
2. Discuss the effects of micro- and nanosized SiC fillers in an epoxy matrix
on the thermal properties of the composite.
3. Discuss the effects of micro- and nanosized SiC fillers in an epoxy matrix
on the mechanical properties of the composite.
4. Discuss the advantages and disadvantages of nanosized SiC fillers on the
fatigue performance of polymeric matrix composites.
REFERENCES
1. Lau, K. and Hui, D., The revolutionary creation of new advanced materials-carbon
nanotube composites, Composites: Part B, 33, 263277, 2002.
2. Becker, O., Varley, R., and Simon, G., Morphology, thermal relaxations and mechan-
ical properties of layerd silicate nanocomposites based upon high-functionality epoxy
resins, Polymer, 43, 43654373, 2002.
3. Agag, T., Koga, T., and Takeichi, T., Studies on thermal and mechanical properties
of polyimide-clay nanocomposites, Polymer, 42, 33993408, 2001.
4. Gianellis, E. P., Polymer layered silicate nanocomposites, Adv. Mater., 8, 2935, 1996.
5. Edelstein, A.S. and Cammarata, R.C., Nanomaterials: Synthesis, Properties and
Applications, Institute of Physics Pub., 1996.
6. Mahfuz, H., Islam, M., Rangari, V., Saha, M., and Jeelani, S., Response of Sandwich
Composites with Nanophased Cores under Flexural Loading, Composites Part B, 35,
543550, 2004.
7. Dagani, R., Putting the nano into composites, Chemical & Engineering News, June 7,
1999.
8. Birringer, R. and Gleiter, H., Encyclopedia of Materials Science and Engineering,
Cahn, R.W. et al., Eds, Suppl. Vol. 1, Pergamon Press, Oxford, 1988, pp. 339.
9. Gleiter, H., Nanocrystalline materials, Prog. Mater. Sci., 33, 223315, 1989.
10. Dagani, R., Nanostructural materials promise to advance range of technologies,
Chemical & Engineering News, November, 1824, 1992.
11. Uyeda, R., Studies of ultrafine particles in Japan: Crystallography. Methods of oper-
ation and technological application, Prog. Mater. Sci., 35, 196, 1991.
12. Karch, J., Birringer, R., and Gleiter, H., Ceramics ductile at low-temperature Nature,
330, 556558, 1987.
13. Teishev, A., Incardons, S., Migliaresi, C., and Marom, G., Polyethylene fibers-
polyethylene matrix composites: Preparation and physical properties, J. Appl. Polym.
Sci., 50, 503512, 1993.
14. Sperling, L.H., Introduction to Physical Polymer Science, 2nd ed., John Wiley & Sons,
New York, 1992.
15. Hlangothi, S.P., Krupa, I., Djokovic, V., and Luyt, A.S., Thermal and mechanical
properties of cross-linked and uncross-linked linear low-density polyethylenewax
blends, Polymer Degradation and Stability, 79, 5359, 2003.
16. Walter, R., Friedrich, K., Privalko, V., and Savadori, A., On modulus and fracture
toughness of rigid particulate filled high density polyethylene, J. Adhesion, 64,
87109, 1997.
17. Rothan, R.N., Mineral fillers in thermoplastics: Filler manufacture and characteriza-
tion, Advanced Polym. Sci., 139, 67107, 1999.
18. Wu, C., Zhang, M., Rong, M., and Friedrich K., Tensile performance improvement
of low nanoparticles filled-polypropylene composites, Composite Sci. Tech., 62,
13271340, 2002.
19. Shiner, C., Timmerman, J., Eboneee, P.M., Williams and Seferis, J., Thermal and
Mechanical Characteristics of Nano Modified Fiber-Reinforced Composites, paper
presented at 48th International SAMPE Symposium, May 1115, 2003, p. 2539.
20. Fukushima, Y., and Inagaki, S., Synthesis of an intercalated compound of montmo-
rillonite and 6-polyamide, J. Inclusion Phenomena, 5, 473482, 1987.
21. Garces, J., Moll, D., Bicerano, J., Fibiger, R., and McLeod, D., Polymeric nanocom-
posites for automotive applications, Adv. Mater. 12, 23, 18351839, 2000.
22. Gilman, J. and Kashiwagi, T., Literature review in Beall, G.W. and Pinnavaia, T., Eds.,
in Polymer-Clay nanocomposites, John Wiley & Sons, New York, 2000, p. 193.
23. Kojima, Y., Usukia, A., Kawasumi, M., Okada, A., Fukushima, Y., Kurauchi, T., and
Kamigaito, O., Mechanical properties of nylon-6/clay hybrid, J. Mater. Res., 8,
11851189, 1993.
24. LeBaron, P., Wang, Z., and Pinnavaia, T., Polymer-layered silicate nanocomposites:
an overview, J. Appl. Clay Sci., 15, 1129, 1999.
25. Lee, A. and Lichtenhan, J., Thermal and viscoelastic property of epoxy-clay and
hybrid inorganic-organic epoxy nanocomposites, J. Appl. Polym. Sci., 73, 19932001,
1999.
26. Eskin, G., Broad prospects for commercial application of the ultrasonic (cavitation)
melt treatment of light alloys, Ulatrasonic Sonochemistry, 8, 319325, 2001.
27. Price, G., Recent developments in sonochemical polymerization, Ultrasonics
Sonochemistry, 10, 277283, 2003.
28. Mahfuz, H., Zaman, K., Hisham, M., Foy, Costee, Haque, A., and Jeelani, Fatigue
life prediction of thick-section S2-glass/vinyl-ester composites under flexural loading,
Transaction of ASME, J. Eng. Mater. Tech., 122, 402408, 2000.
29. Mahfuz, H., Mamun, W., and Jeelani, S., High strain rate response of sandwich
composites: Effect of core density and core-skin debond, J. Adv. Mater., 34, 1, 2226,
2002.
30. Vaidya, U., Kamath, M., Hosur, M., Mahfuz, H., and Jeelani, S., Manufacturing and
low velocity impact response of sandwich composites with hollow and foam filled
Z-pin reinforced core, J. Composites Tech. Res., 21, 2, 8497, 1999.
31. Vijaya Kumar, R., Koltypin, Y., and Gedanken, A., Preparation and characterization
of nickel-polystyrene nanocomposite by ultrasound irradiation, J. Appl. Polym. Sci.,
86, 160, 2002.
32. Vijaya Kumar, R., Koltypin, Y., Cohen, Y.S., Cohen, Y., Aurbach, D., Palchik, O.,
Felner, I., and Gedanken, A., Preparation of amorphous magnetic nanoparticles embed-
ded in polyvinyl alcohol using ultrasound radiation, J. Mater. Chem., 10, 1125, 200.
33. Mahfuz, H., Maniruzzaman, M., Vaidya, U., and Jeelani, S., Fatigue damage and
effects of stress ratio on the fatigue life of carbon-carbon composites, Theor. Appl.
Fracture Mech., 24, 2131, 1995.
34. Mahfuz, H., Partha, S.D., Jeelani, S., Baker, D. M., and Johnson, S., Effect of
missioncycling on the fatigue performance of SiC coated carbon-carbon composites,
Int. J. Fatigue, 15, 4, 283291, 1993.
35. Mahfuz, H., Maniruzzaman, M., Vaidya, U., and Jeelani, S., Response of SiC/Si3N-
4 composites under cyclic loadingan experimental and statistical analysis, J. Eng.
Mater. Technol., 119, 186193, 1997.
36. Mahfuz, H., Adnan, A., Rangari, V., and Jeelani, S., Carbon nanoparticles/whiskers
reinforced composites and their tensile response, Composites Part A, Appl. Sci.
Manufacturing, 35, 519527, 2004.
37. Mahfuz, H., Rangari, V., Islam, M., and Jeelani, S., Fabrication, synthesis and
mechanical characterization of nanoparticles infused polyurethane foams, Composites
Part A, Appl. Sci. Manufacturing, 35, 453460, 2004.
10
Synthesis and
Characterization of Metal-
Ceramic Thin-Film
Nanocomposites with
Improved Mechanical
Properties
Dhanjay Kumar, Jagannathan Sankar, and
Jagdish Narayan
CONTENTS
10.1 Introduction
10.2 Theory of Pulsed Laser Deposition
10.4.1 Mechanical Properties
10.4.2 Structural Characterization
10.5 Conclusions
Problems
Acknowledgments
References
10.1 INTRODUCTION
Composite materials consisting of metallic clusters or crystals of nanometer dimen-
sion embedded in an insulator host can exhibit special optical, electrical, magnetic,
and mechanical properties that may be used in various technological applications.
18
For example, metal nanocrystals embedded in insulating materials offer enhanced
mechanical properties, making them promising candidates for use in machine
tools.
914
The search for materials with enhanced hardness is driven by both the
scientific curiosity of researchers to explore the possibilities of synthesizing a mate-
rial whose hardness could approach or even exceed that of diamond and the technical
importance of hard materials for wear protection, e.g., of machine tools. The impor-
tance of hard-wear protective coatings for machining applications is illustrated by
the fact that today more than 40% of all cutting tools are coated by wear-resistant
coatings and the market is growing fast. Wear-resistant hard coatings for high-speed
dry machining would allow the industry to increase the productivity of expensive
automated machines and to reduce the high costs presently associated with the use
of environmentally hazardous coolants.
The metal nanocrystals embedded in insulating materials have been synthesized
by quenching and heat treatments,
15
sol-gel processes,
16
sputtering,
17
and ion implan-
tation.
18
Most of these processes are multistep processes, in which a postdeposition
treatment is often needed to optimize the properties. However, these treatments can
also alter the average grain size, size distribution, and spatial arrangement of the
nanocrystals, with possible unfavorable effects on the mechanical properties of the
nanocomposites. Unlike these methods, multiple-target sequential pulsed laser dep-
osition (PLD) permits independent control of synthesis of the nanocrystals and the
embedding matrix. PLD has shown particular success in stoichiometric thin-film
deposition of complex oxides. Energetic (!100 eV) ions produced by ablation yield
smooth, high-density films with good adhesion, especially desirable for mechanical
applications. In this chapter, we report the synthesis of alternating-target PLD of Fe
and Ni nanocrystals embedded in an Al
2
O
3
matrix. The material produced is a thin-
film composite consisting of metallic nanocrystals embedded in an insulator host.
The structural quality of the thin-film composite was evaluated using high-resolution
transmission electron microscopy and scanning transmission electron microscopy
with atomic number contrast. This revealed the formation of a biphase system with
thermodynamically driven segregation of Ni and alumina during pulsed laser depo-
sition. This new nanocomposite material exhibits superior mechanical properties with
enhanced hardness and Youngs modulus. The improvement in hardness of Al
2
O
3
thin films by embedding metal nanocrystals is related to the evolution of a micro-
structure that efficiently hinders the manipulation and movement of dislocations and
the growth of microcracks. This is achieved by grain boundary hardening.
10.2 THEORY OF PULSED LASER DEPOSITION
The pulsed laser deposition technique has been used to prepare the thin-film nano-
composite samples discussed in the present chapter. Since this is a relatively new
technique with respect to other thin-film deposition techniques such as evaporation,
sputtering, and molecular beam epitaxy, it is appropriate to briefly describe the theo-
retical aspects of pulsed laser deposition. The pulsed laser deposition (PLD) process
was first used more than two decades ago.
19
However, it gained prominence recently
when it was found to be the most convenient and efficient technique for the synthesis
of new high-temperature superconducting thin films.
20
The schematic of a pulsed laser
deposition chamber is shown in Fig. 10.1. In a typical PLD of thin films, a pulsed
laser strikes a solid bulk target. Some of the target materials are removed, escaping in
the form of a plume. Part of the plume comes in contact with the surface of a heated
substrate kept a few centimeters away from the target. The plume, consisting of the
building block (lattice) of the material, covers the substrate. The result is the fabrication
of a thin film of the given material with the same chemical structure as the target. PLD
offers numerous advantages, including film stoichiometry close to the target, a low
contamination level, a high deposition rate, and nonequilibrium processing. The rela-
tively easily accessible experimental parameters in PLD make it very attractive for the
synthesis of high-temperature superconducting thin films. These parameters are the
substrate temperature, the energy of the atom flux, the relative and absolute arrival
rate of atoms for compound films, and the pressure in the chamber. In order for any
thin-film technique to be highly suitable for the growth of thin-film materials, the
following conditions must be fulfilled: (1) the substrate temperature should be accu-
rately controlled; (2) all the atoms in the deposition flux should have an energy of
5 eV to 10 eV to promote surface diffusion, nucleation, and high sticking probability
without producing damage; (3) the atoms of the multicomponent material should arrive
with the precise relative abundance required for the compound; and (4) it is useful to
do all the processing in a high partial pressure of reactive gases such as oxygen to
maintain stoichiometry of volatile species on the heated substrates and to control the
energy of the deposited flux. The plasma generated by laser irradiation of the target
can be used to assist the growth of very high quality thin films at relatively low
processing temperatures. The processing temperature in PLD is generally 50100C
lower than used in other thin-film growth techniques. Atomically sharp multilayer
structures and superlattices can easily be fabricated by this technique. These advantages
have made the PLD method one of the most popular methods to deposit thin films.
The basic experimental design for thin-film deposition by laser ablation is similar
to any other physical vapor deposition process. The apparatus includes a vacuum
chamber, a substrate holder with precise temperature control, and source materials
(target). Usually an excimer laser that uses a mixture of Kr, F
2
, He, and Ne generates
pulses with wavelengths varying from 193 nm to 308 nm is employed for target
FIGURE 10.1 Schematic of a pulsed laser deposition system.
Quartz lens
Vacuum
chamber
Pump
Heater
Substrate
Target
Plume
Traget carousel
Target holder
Laser beam
Motor
248 nm
KrF laser
ablation. Ultraviolet lasers have been found to result in congruent ablation that varies
from 25 to 30 ns duration and the energy density ranges from 1.0 to 2.5 J/cm
2
. An
aperture is placed across the beam so that the nonuniform edge effects can be
minimized. The laser fluence is varied by either varying the laser output energy or
by focusing the beam. The optical imaging system consists of a mirror with reflective
coating and a planoconcave lens. The laser beam is admitted to the chamber via a
quartz window that is susceptible to damage if the energy density is too high or if
the window is covered by dust particles. The sides of the chamber contain several
quartz windows. One is used to admit the laser beam, while the others are used for
in situ plasma diagnostics and for monitoring the growth. The chamber contains a
sample heater block, a rotating target holder, and a shutter. The target block is
mounted on a linear motion feedthrough that is used to vary the target-substrate
distance. The target is rotated at about 1 Hz, which is incommensurate with the laser
pulse rate but provides a more uniform erosion of the target. The oxygen pressure
during deposition is regulated using a mass flow controller and a throttle valve.
Although PLD appears to be an evaporation process, the stoichiometry and the
nature of species are quite different from standard evaporation processes. This is due
to the complex interaction of the laser beam with the target and plasma and subsequent
deposition of species. Depending on the type of interaction of the laser beam with the
target, the PLD process can be classified into four separate regimes: (1) interaction of
the laser beam with the target material resulting in ablation of the surface layers;
(2) interaction of the ablated material with the incident laser beam resulting in an
isothermal plasma formation and expansion; (3) anisotropic adiabatic expansion of the
plasma leading to the characteristic nature of the laser deposition process; and (4)
interaction of the plasma with the background gas. A schematic view of the interaction
processes is shown in Fig. 10.2. The first two regimes (marked (a) and (b) in Fig. 10.2)
FIGURE 10.2 Schematic diagram showing the different phases present during laser irradia-
tion of a target: (a) unaffected target, (b) evaporated target, (c) dense plasma absorbing laser
radiation, and (d) expanding plasma outer edge transparent to the laser beam.
Laser beam
(a) (b) (c) (d)
X
Bulk target Plasma cloud
start with the laser pulse and continue throughout the laser pulse duration, while the
next two regimes initiate after the laser pulse terminates. Under PLD deposition con-
ditions, where the pulse energy density is in the range of 1 to 10 J/cm
2
, the target
ablation can be considered to be thermal in nature, while the interaction of the laser
beam with the evaporated material gives rise to thermal characteristics of the species
in the plasma.
The thermal effects of pulsed nanosecond laser irradiation of materials are
determined by the laser pulse parameters (temporal power density I(t), pulse duration
t
p
, wavelength, etc.), optical properties (reectivity R, absorption coefcient ), and
the thermal properties of the material (thermal conductivity K, latent heat per unit
volume L
v
, specic heat per unit volume C
v
, ablation temperature T
v
, etc.). The
thermal diffusivity D = K/C
v
denes the thermal diffusion length (2Dt
p
)
0.5
, which
determines the spread of the temperature prole during the laser pulses. The tem-
perature in the target T(x,t) during laser irradiation is controlled by the heat ow
equation given by
23
(10.1)
with appropriate boundary conditions that take into account the formation and move-
ment of the solid-liquid (or liquid-vapor) interface. Here, x refers to the direction in
the plane perpendicular to the target, and t refers to the time. The second term on the
right-hand side of the equation is the heat generation term due to the absorption of
the incident laser beam by the target. If the surface of the material is highly absorbing
( 10
6
cm
1
) to the incident laser beam, the heat generation term can be removed
from Equation 10.1, and applied to the front surface boundary condition.
The laser-target interactions play a critical role in the lm quality deposited by
the PLD technique. Depending on the nature of the heating and thermophysical
properties of the materials, two separate regimes can be distinguished
24
: (1) a surface
heating regime in which the optical absorption depth is much smaller than the thermal
diffusion distance, (2Dt
p
)
0. 5
>> 1, and (2) a volume heating regime in which the
optical absorption depth is much larger than the thermal diffusion distance,
(2Dt
p
)
0.5
<< 1. The two regimes are shown schematically in Fig. 10.3. In the rst
case, the laser energy is absorbed in the surface layer, and the thermal diffusivity
controls the heating and ablation characteristics. During ablation of the target, a
planar vaporization interface may initiate from the surface and propagate into the
bulk of the target. However, at higher power densities, subsurface heating effects
that may lead to explosive removal of material from the target become important,
and thus lead to nonlinear ablation effects.
25
The transformation from surface to
volume ablation regimes is dependent on several factors such as incident power
density, absorption coefcient, and the thermophysical properties of the target. In
the volume heating regime, the thermal conductivity plays an insignicant role in
controlling the laser-target interactions, and the ablation depth is determined mainly
by the optical absorption depth, which is given by the inverse of the absorption
coefcient. In this regime, internal heating subsurface effects may predominate,
leading to removal of particles from the target, and these particles may shield the
C T
T x t
t x
K t
T x T
x
R
v
( )
( )
( )
( )
+ ( )
, ,
1 II x t e
x
0
, ( )

target from the incoming laser beam. The term subsurface superheating has been
developed to describe temperature inversion on the surface during laser ablation. In
the case of planar surface ablation when the vaporization interface moves rapidly
into the target, the surface is constantly being cooled due to the latent heat of ablation.
However, the absorption of the laser beam in the target is characterized by a
finite absorption depth in the material. The subsurface layers thus are heated directly
by the laser beam where the heat dissipation mechanism is due to the thermal
conduction losses in the target. As the energy density is increased, the subsurface
temperatures may vastly exceed the surface temperature, which corresponds to the
ablation point of the material (Fig. 10.4). These subsurface superheating heating
effects may lead to explosive removal of material from the target in the form of
particles. These particles may in turn shield the laser beam from the target, thus
giving rise to nonlinear ablation characteristics.
According to Singh et al.,
26
the time after which the ablation front may change
from a planar to volume type will depend on the degree of internal superheating
and the time required for the nucleation to form a subsurface gaseous phase. Because
of the number of unknowns, the nucleation of the stable gaseous phase near the
FIGURE 10.3 Schematic of optical absorption and thermal diffusion profiles for (a) the
surface heating regime where the optical absorption length is very small and insignificant
compared to the thermal diffusion length, and (b) the volume heating regime where the optical
absorption length is much greater than the thermal diffusion length.
Thermal
diffusion
distance
Thermal
diffusion
distance
Absorption
depth
Absorption depth
Temperature
profile
Temperature
profile
(a)
(b)
surface cannot be quantitatively analyzed using the basic nucleation theory. However,
it is known that the nucleation rate is a strong function of the temperature, and thus
a larger degree of superheating increases the probability of more rapid volume
nucleation of the material. Using this approach, Singh et al.
26
analyzed the effect of
various processing factors on subsurface superheating.
The high surface temperature induced by laser irradiation leads to emission of
positive ions and electrons from a free surface. The thermionic emission of positive
ions and electrons from hot surfaces has been widely recognized in the literature.
27
The flux of ions and electrons as a function of temperature can be predicted by
Richardson's and Langmuir-Sahas equations, respectively. Both of these equations
show an exponential increase in the fraction of ionized species with temperature;
higher ionized fractions than predicted by the Langmuir-Saha equation have been
observed in laser-irradiated targets.
28
This has been attributed to the higher temper-
atures induced by absorption of the laser beam by the evaporating material and
electron impact ionization.
The physical mechanisms involved in the absorption and reflection of the laser
energy by the evaporating material were identified in the early experiments as the
sources for very high temperature (a1 KeV) plasma. The penetration and absorption
of the laser beam by the plasma depend on the electron-ion density, temperature,
and the wavelength of the laser light. For penetration (or reflection) of the incident
laser beam, the plasma frequency (v
p
) should be lower than the laser frequency. For
excimer laser wavelengths (O 193308 nm), the laser frequency varies from 9.74 u
10
14
cm
1
to 1.5 u l0
15
cm
1
. The plasma frequency is given by
(10.2)
where n
e
is the electron concentration of the plasma. Using Equation 10.2, the critical
density for reflection of the laser beam is determined to be 13 u 10
22
/cm
3
. This
critical density is approximately equal to the concentration of atoms in the solid or
liquid. The presence of a diffuse plasma boundary and a gradual decrease in the
plasma density away from the surface results in a lower electron density than the
FIGURE 10.4 Schematic of subsurface heating of the target during laser ablation.
Target Target Target
Evaporation onset
Middle of
pulse
Middle of
pulse
Ablated
thickness
Ablated
thickness
T
e
m
p
e
r
a
t
u
r
e
T
evap
v n
p e
= 8 9 10 0 5
3
. .
estimated value. Thus, energy losses due to the reflection of laser light from the
plasma are insignificant.
The material that evaporated from the hot target is further heated by the absorp-
tion of the laser radiation. The primary absorption mechanism for ionized plasma
is electron-ion collision. The absorption primarily occurs by an inverse Bremmstrahl-
ung process, which involves the absorption of a photon by a free electron. The
absorption coefficient (D
plasma
) of the plasma can be expressed as
(10.3)
where Z, n
i
, and T, are the average charge, ion density, and temperature of the plasma,
respectively, and h, k, and v are the Planck constant, Boltzmann constant, and frequency
of the laser light, respectively. The laser energy is highly absorbed if (D
p
X) a 1, where
X(t) is the dimension perpendicular to the target of the expanding plasma. This equation
shows that the absorption coefficient of the plasma is proportional to Thus the
plasma absorbs the incident laser radiation only at distances very close to the target
where the densities of the charged particles are very high.
Nanocrystalline nickel and iron crystallites were embedded in an alumina matrix using
a KrF excimer laser (252 nm, 30 ns full width at half maximum) focused alternately
onto high-purity targets of nickel or iron and alumina. The depositions were carried
out on silicon substrates in a high-vacuum environment (a5 u 10
7
torr). The substrate
temperature was approximately 500qC. The energy density and repetition rate of the
laser beam used were 2 J/cm
2
and 10 Hz, respectively. The size distribution of metallic
particles and the crystalline quality of both the matrix and metallic particles were
investigated by cross-sectional scanning transmission electron microscopy with atomic
number (Z) contrast (STEM-Z). Figure 10.5 is a schematic diagram showing the
incoherent imaging under STEM conditions in which a high-angle annular detector is
employed to collect the incoherent tail of the scattered electron distribution. The image
is formed by focusing a very fine probe (2.2 at 100 KeV) on the sample. There is
no imaging lens in this system. Instead, high-angle scattering is detected as a function
of probe position and used to map the scattering power of the sample, which depends
on the square of the atomic number. This is a dark-field technique, so that the atomic
columns are seen as bright spots in the image; the heaviest columns are always
brightest, allowing unambiguous discrimination of atoms with different atomic num-
bers. The diagram shows that this technique is able to resolve Ga and As in GaAs,
even though these two atoms have a small atomic number difference between them.
21,22
It should be noted that the STEM-Z technique is capable of resolving the location of
oxygen atoms at the interface. This is accomplished by simultaneously using the EELS
mode in which a low-angle detector collects the oxygen signal. Thus, with the simul-
taneous use of the STEM-Z and EELS, both low- and high-atomic-number atoms can
be identified at the particle/oxide interface.
Depth-sensing nanoindentation continuous stiffness measurements (CSMs) were
made on the nanocomposite thin films. A MTS Nanoindenter-XP with a Berkovich
p i
Z n T v hv kT =
( )
( )
3 69 10 1
8 3 2 0 5 3
. / exp /
.
n
i
2
.
triangular pyramidal tip was used for these measurements. The instrument monitored
and recorded the dynamic load and displacement values during the indentation. A
series of arrays of indentations were made at a constant strain rate of 0.05 s
1
and at
depths of 750 nm for hardness and modulus measurements. Four samples were made
per deposition and three indentation tests were run on each sample. Each test comprised
an array of six indents. Load displacement data collected from the indentations were
used to calculate the hardness and Youngs modulus using Oliver and Pharr analysis.
10.4.1 MECHANICAL PROPERTIES
Shown in Fig. 10.6 is the variation of hardness of Ni-Al
2
O
3
thin-film composites as
a function of the size of the Ni particles embedded in the Al
2
O
3
matrix. For the sake
of comparison, we have shown in the figure the hardness data of pure alumina film
deposited under identical conditions. It is clear from the figure that the hardness of
the Ni-alumina composite is significantly higher than that of pure alumina films.
However a further increase in the particle size of the embedded Ni particles resulted
FIGURE 10.5 Schematic showing the optical arrangement for atomic imaging and analysis with
the scanning transmission electron microscope. The inset shows the Z-contrast image of GaAs.
Annular
detector
Spectrometer
CCD-EELS
detector
2.2 or 1.3
incident probe
scanned across
specimen
Z-contrast image of GaAs
Ga As
1.4
in the deterioration of hardness, suggesting that there is an optimum size of metal
particle that produces the greatest hardness in the composite. Supporting evidence
in favor of this suggestion comes from hardness measurements performed on different
thin-film composite systems consisting of Fe nanocrystals in alumina thin films.
The results obtained are shown in Fig. 10.7. According to the results in this figure,
the hardness of the first Fe-Al
2
O
3
composite (size of Fe particles is 3 nm) is in fact
lower than that of pure alumina film. However, the hardness of the Fe-Al
2
O
3
FIGURE 10.6 Hardness as a function of Ni nanoparticle size in different samples of the Ni-
Al
2
O
3
thin-film composite system.
FIGURE 10.7 Hardness as a function of Fe nanoparticle size in different samples of the Fe-
Al
2
O
3
thin-film composite system.
0 4 8 10 12
18
20
22
24
26
28
H
a
r
d
n
e
s
s

(
G
P
a
)
Particle size (nm)
2 6
0
16
20
24
28
H
a
r
d
n
e
s
s

(
G
P
a
)
Particle size (nm)
8 6 4 2
composite becomes greater than that of pure alumina thin film when the size of
the embedded Fe nanocrystals increases to 4.5 nm. The hardness reaches a peak at
7 nm, beyond which the hardness of the Fe-Al
2
O
3
thin-film composite (Fe particle
size is 9 nm) is again lower than that of pure alumina thin film.
The improvement in values of hardness of Al
2
O
3
thin films by embedding metal
nanocrystals is related to the evolution of a microstructure that efficiently hinders
the manipulation and movement of dislocations and the growth of microcracks,
which in turn is achieved by grain boundary hardening as described by the Hall-
Petch relationship, which is valid down to a crystallite size of 2050 nm
2931
:
(10.4)
Here V
c
is the critical fracture stress, d is the crystallite size, and V
0
and k
gb
are constants.
With the crystallite size decreasing to this limit, the fraction of the material in the grain
boundaries increases, which leads to a decrease in material strength and hardness due
to increased grain boundary sliding.
3236
A simple phenomenological model has been
used to describe the softening in terms of an increasing volume fraction of the grain
boundary material f
gb
with the crystallite size decreasing below 16 nm
37
:
(10.5)
with f
gb
D(1/d). Due to the flaws present, the hardness of grain boundary materials
h
gb
is smaller than that of the crystallites H
c
. Thus the average hardness of the material
decreases with d decreasing below 10 nm, an effect commonly known as the reverse
Hall-Petch effect.
38
Recent computer simulation studies confirm that the reverse
Hall-Petch dependence in nanocrystalline materials is due to the grain boundary
sliding that occurs due to a large number of small sliding events on atomic planes
at the grain boundaries, without thermal activation. This will ultimately impose a
limit on how strong the nanocrystalline metal may become.
39,40
Although many
details are still not understood, there is little doubt that grain boundary sliding is
the reason for softening in this crystallite size range. Therefore a further increase of
the strength or hardness with decreasing crystallite size can be achieved only if grain
boundary sliding is blocked by appropriate design of the material.
40,41
We have tried
to achieve this condition by having a uniform distribution of isolated nanocrystals
in an amorphous thin-film matrix. In such situations, the grain boundaries formed
at the interface are very strong and avoid or greatly reduce the grain boundary sliding.
We believe that this effect is primarily responsible for the higher values of hardness
for metal-alumina thin-film composites composed of isolated nanocrystalline metal
particles (size below 10 nm) in amorphous matrices.
10.4.2 STRUCTURAL CHARACTERIZATION
Shown in Fig. 10.8 is the cross-sectional STEM-Z micrograph of Ni nanocrystallites
embedded in an alumina matrix. The Z-contrast in the STEM provides an incoherent
image in which changes in atomic structure and composition across an interface can
be interpreted directly without the need for preconceived atomic structure models.

c
gb
k
d
= +
0
H f f H f H
gb gb c gb gb
( ) ( ) = + 1
This is true provided the incident electron probe is smaller than the lattice spacing
(for a sample oriented along a major axis) or atomic columns. Since the probe is
scanned, the resultant image is a map of the column scattering power, which in turn
depends on the atomic number Z
2
of each column. The spatial resolution is limited
primarily by the probe size, which is 1.2 in our field emission instrument. The
Z-contrast image in the STEM is formed by electrons scattered through high angles.
The average size of the Ni nanocrystals was determined to be ~5 nm. It is clear from
the cross-sectional images that the Ni nanoparticles are mostly uniform in size and are
well separated from each other. The average intralayer separation between the particles
is estimated to be ~25 nm. In order to evaluate the crystalline quality of the metal
particles and the alumina matrix, a high-resolution STEM-Z was recorded (Fig. 10.9)
from a single metal particle and the alumina matrix in its immediate vicinity. It is clear
FIGURE 10.8 STEM-Z image of the Ni nanoparticles embedded in the alumina matrix.
FIGURE 10.9 A STEM-Z image of a single Ni nanoparticle and the adjoining alumina matrix.
from this figure that the Ni nanoparticle is crystalline and the alumina matrix is amor-
phous in nature. Such a biphase system is formed as both materials are immiscible (i.e.,
they display thermodynamically driven segregation during deposition) and the cohesive
energy at the interface between both phases is high. The amorphous phase of the matrix
can provide high structural flexibility in order to accommodate the coherency strain
without forming dangling bonds, voids, or other flaws as observed in Fig. 10.9. In this
case, the nanostructure would be stable and grain boundary sliding is greatly reduced.
The improvement in the values of hardness of Al
2
O
3
by embedding Ni and Fe
nanoparticles may also be explained using the Koehler concept of multilayers for
the design of strong solids.
42
According to this concept, under applied stress a
dislocation that would form in a softer layer (metal in the present study) would move
toward the metal-ceramic interface, and elastic strain in the second layer (alumina
in the present study) with the higher elastic modulus would cause a repulsive force
that would hinder the dislocation from crossing that interface. Therefore the hardness
of Ni-Al
2
O
3
and Fe-Al
2
O
3
thin-film composites, which are essentially multilayered
structures consisting of continuous layers of alumina thin films and discontinuous
films of metals, is much stronger than expected from the rule of mixtures.
10.5 CONCLUSIONS
In summary, thin-film composite materials consisting of metallic nanocrystals embed-
ded in an insulator were fabricated. These films exhibited significantly improved
mechanical properties. The best hardness values measured using a nanoindentation
technique were 2030% larger than the hardness of pure alumina films fabricated under
identical conditions. The improvement in the hardness of Al
2
O
3
thin films by embedding
metal nanocrystals is related to the evolution of a microstructure that impedes the
manipulation and movement of dislocations and the growth of microcracks. The imped-
ance in grain boundary movement is brought about by grain boundary hardening due
to formation of well-separated metallic nanocrystallites in amorphous matrices.
PROBLEMS
1. Describe the role of the pulsed laser deposition method in the fabrication
of thin-film nanocomposite materials.
2. Discuss the mechanism for the improvement in ceramic thin films by the
addition of nanodimensional metal particles.
3. Explain how the mechanical properties of hard materials such as alumina
can be improved by the addition of soft materials such as nickel or iron.
ACKNOWLEDGMENTS
The authors would like to thank Dr. S.J. Pennycook of Oak Ridge National Laboratory
for collaboration on the STEM-Z characterization. Partial funding for this research
was obtained from a U.S. National Science Foundation nanoscale exploratory research
(NER) grant and from the Center for Advanced Materials and Smart Structures.
REFERENCES
1. Stamm, C., Marty, M., Vaterlaus, A., Weich, V., Egger, S., Maier, U., Ramsperger,
U., Fuhrmann, H., and Pescia, D., Science, 282, 449, 1998.
2. Lu, L., Sui, M.L., and Lu, L., Science, 287, 1463, 2000.
3. Timp, G., Ed., Nanotechnology, Springer, New York, 1999.
4. Kumar, D., Zhou, H., Nath, T.K., Kvit, A.V., and Narayan, J., Appl. Phys. Lett., 79,
2817, 2001.
5. Kumar, D., Narayan, J., Kvit, A.V., Sharma, A.K., and Sankar, J., J. Mag. Mater.,
232, 161, 2001.
6. Prinz, G.A., Science, 282, 1660, 1998.
7. Mehn, M., Punadjela, K., Bucher, J., Rousseausx, F., Decanini, D., Bartenlian, B.,
and Chappert, C., Science, 272, 1782, 1996.
8. Katiyar, P., Kumar, D., Nath, T.K., Kvit, A.V., Narayan, J., Chattopadhyay, C.,
Gilmore, W.M., Coleman, S., Lee, C.B., Sankar, J., and Singh, R.K., Appl. Phys.
Lett., 79, 1327, 2001.
9. Ballesteros, J.M., Serna, R., Solis, J., Afonso, C.N., Petford-Long, A.K., Osborne,
D.H., and Haglund, Jr., R.F., Appl. Phys. Lett., 71, 2445, 1997.
10. Hirono, S., Umemura, S., Tomita, M., and Kaneko, R., Appl. Phys. Lett., 80, 425,
2002.
11. Veprek, S., J. Vac. Sci. Technol. A, 17, 2401, 1999.
12. Cselle, T. and Barimani, A., Surf. Coat. Technol., 7677, 712, 1995.
13. Rodeghiero, E.D., Tse, O.K., Chisaki, J., and Giannelis, E.P., Mater. Sci. Eng. A, 195,
151, 1995.
14. Sekino, T., Nakajima, T., Ueda, S., and Niihara, K., J. Am. Ceram. Soc., 80, 1139,
1997.
15. Lu, K., Wang, T., and Wei, W.D., J. Appl. Phys., 69, 522, 1991.
16. De, G., Tapfer, L., Catalano, M., Battaglin, G., Caccavale, F., Gonella, F., Mazzoldi,
P., and Haglund, Jr., R.F., Appl. Phys. Lett., 68, 3820, 1996.
17. Liao, H.B., Xiao, R.F., Fu, J.S., Yu, P., Wong, G.K.L., and Sheng, P., Appl. Phys.
Lett., 70, 1, 1997.
18. Mazzoldi, P., Arnold, G.W., Battaglin, G., Bertoncello, R., and Gonella, F., Nucl.
Instrum. Methods Phys. Res. B, 91, 478, 1994.
19. Chrisey, D.B. and Hubler, G.H., Pulsed Laser Deposition of Thin Films, John Wiley
& Sons, New York, 1994.
20. Singh, R.K. and Kumar, D., Mater. Sci. Eng., R22, 113185, 1999.
21. Mcgibbon, A.J., Pennycook, S.J., and, Angelo, J.E., Science, 269, 519, 1995.
22. Pennycook, S.J., Ann. Rev. Mater. Sci., 22, 171, 1992.
23. Singh, R.K. and Narayan J., Mater. Sci. Eng., B3, 2923, 1989.
24. Bloemerben, N., in Ferris, S.D., Leamy, H.J., Poate, J.M., Eds., Laser Solid Interac-
tions, American Institute of Physics, New York, 1979, p. 1.
25. Singh, R.K., Bhattacharya, D., and Narayan, J., Appl. Phys. Lett., 57, 2022, 1990.
26. Singh, R.K., J. Non-Crys. Solids, 178, 199, 1994.
27. Reddy, J.F., Effects of High Power Laser Radiation, Academic Press, New York, 1971.
28. Nakayama, T., Okigawa, M., and Itoh, N., Nucl. Instru. Meth., B1, 301, 1984.
29. Hertzberg, R.W., Deformation and Fracture Mechanics of Engineering Materials,
3rd ed., John Wiley & Sons, New York, 1989.
30. Hall, E.O., Proc. Phys. Soc. London, Sect. B64, 747, 1951.
31. Petch, N.J., J. Iron Steel Inst., London, 174, 25, 1953.
32. Siegel, R.W. and Fougere, E., Mater. Res. Soc. Symp. Proc., 362, 219, 1995.
33. Siegel, R.W. and Fougere, G.E., Nanostruc. Mater, 6, 205, 1995.
34. Hahn, H. and Padmanabhan, K.A., Philos. Mag. B, 76, 559, 1997.
35. Yip, S., Nature, 391, 532, 1998.
36. Schiotz, J., Di Tolla, E., and Jacobsen, K.W., Nature, 391, 561, 1998.
37. Carsley, J.E., Ning, J., Milligan, W.M., Hackney, S.A., and Aifantis, E.C., Nanostruc.
Mater., 5, 441, 1995.
38. Chokshi, A.H., Rosen, A., Karch, J., and Gleiter, H., Scr. Metall., 23, 1679, 1989.
39. Nieman, G.W., Weertman, J.R., and Siegel, R.W., J. Mater. Res., 6, 1012, 1991.
40. Sanders, P.G., Youngdahl, C.J., and Weertman, J.R., Mater. Sci. Eng. A, 234, 77, 1997.
41. Youngdahl, C.J., Sanders, P.G., Eastman, J.A., and Weertman, J.R., Scr. Mater., 37,
809, 1997.
42. Koehler, J.S., Phys. B, 2, 547, 1970.
11
Macroscopic Fibers
of Single-Walled
Carbon Nanotubes
Virginia A. Davis and Matteo Pasquali
CONTENTS
11.1 Introduction
11.2 Fibers Produced Directly from SWNT Synthesis
11.3 Electrophoretic Spinning
11.4 Conventional Fiber Spinning
11.4.1 Melt-Spun Composite Fibers
11.4.2 Solution-Spun SWNT Fibers
11.4.2.1 SWNT/Liquid Crystalline Polymer Composite
Fibers
11.4.2.2 Fibers Produced from SWNT/Surfactant
Dispersions
11.4.2.3 Fibers Produced from SWNT/Superacid
Dispersions
11.5 Conclusion
Problems
Acknowledgments
References
11.1 INTRODUCTION
On the nano- to microscale, theoretical calculations and experimental measurements
indicate that single-walled carbon nanotubes (SWNTs) have electrical conductivity
and current-carrying capacity similar to copper,
1,2
thermal conductivity higher than
diamond,
3,4
and mechanical strength higher than any naturally occurring or man-
made material.
5,6
Although precise numbers are hard to pinpoint because of the
limited accuracy of theoretical calculations and experiments on small bundles of
nanotubes (NTs), Baughman et al. cite ballistic conductance in metallic SWNTs, a
theoretical thermal conductivity of 2000 W/m K, an elastic modulus of 640 GPa,
and a tensile strength of 37 GPa.
7
Some microelectronic and materials character-
ization applications (e.g., an atomic force microscopy tip) may be able to utilize the
properties of individual and small groups of SWNTs. However, many important
applications require low-cost synthesis of SWNTs and manufacturing methods that
preserve some or all of the nanoscale properties in a macroscopic object.
8,9
Whereas
at least one solution to the economic mass production of SWNTs is now available
through the HiPco process
10
and potentially by other scalable routes,
11,12
technologies
for processing SWNTs into macroscopic materials are at a much earlier stage. Some
of the most promising ones for manufacturing SWNT bers are reviewed here.
Macroscopic carbon nanotube bers have the potential to form high-strength,
lightweight, thermally and electrically conducting structural elements at lower cost
than other forms of SWNTs such as unaligned bucky paper.
13
Whereas futuristic
applications such as space tethers and the space elevator
14
will require ultra-strong
SWNT bers, many other applications will require supplementary multifunctional
properties and not such high mechanical strength.
15
The electrical properties may
be used for highly efcient transmission of electricity over long distances. Thermal
properties could be exploited in microelectronic applications where thermal man-
agement is an increasing problem as miniaturization progresses.
14
The extremely
high surface area of SWNTs (above 300 m
2
/g) could be exploited in carbon nano-
tube supercapacitors, electrochemically driven articial muscles, hydrogen storage,
and electrical energy harvesting.
15
Carbon nanotube ber actuators have already
been tested at temperatures in excess of 350C with a maximum observed actuator
stress of 26 MPa, which is roughly 100 times higher than natural muscle.
7
Potential
uses include robots for planetary exploration and controlling blade pitch in jet
engine rotors at temperatures above 1000C.
16
Fibers containing SWNTs have been produced by four main types of processes:
solid-state production or growth, electrophoretic processing, melt spinning, and
solution spinning i.e., liquid-state processing. Melt and solution spinning seem
to be the most viable techniques for commercial scale-up; moreover, continuous
bers consisting predominantly or solely of SWNTs have been produced only by
solution spinning. This chapter emphasizes liquid-state processing and only briey
summarizes other techniques.
11.2 FIBERS PRODUCED DIRECTLY
FROM SWNT SYNTHESIS
Fibers of carbon nanotubes micrometers to millimeters long have been produced
by variations of chemical vapor deposition (CVD).
8,17,18
For example, lengths up to
20 m have been achieved using 1,3-dicyclohexylcarbodiimide to polymerize oxidized
SWNTs into strands 50 to 150 nm in diameter.
19
Ribbons with preferential alignment
(50140 m wide, 440 m thick and 100 mm in length) have also been produced
by heating oxidized acid-treated tubes at 100C. Graphitization of the ribbons at
2200C for 2 h under argon with a pressure of 0.5 MPa resulted in an increase of
ribbon density from 1.1 to 1.5 g/cm
3
and an increase in Youngs modulus from 24
to 60 GPa.
20
Electrical resistivity of as-grown ribbons was roughly 1 mcm at room
temperature.
On this microscale, alignment during tube growth can be enhanced by electric
elds. Aligned SWNTs on the order of 10 m can be grown under an electric eld
in the range of 0.52 V/m.
21
Aligned CVD-grown nanotubes have been self-
assembled into yarns up to 30 cm in length by drawing them out of arrays several
hundred micrometers in height in a process similar to drawing silk out of a cocoon;
22
an array of 1 cm
2
area and 100 m height could generate about 10 m of yarn. The
diameter of the yarn can be controlled by the size of the drawing tool, with smaller
tools generating smaller diameter yarns. The yarns were used to construct a light
bulb lament and a polarizer suitable for use in the ultraviolet region.
Zhu et al. produced aligned strands up to 20 cm in length and several hundred
micrometers in diameter using a oating catalyst method in a vertical furnace where
n-hexane was catalytically pyrolyzed in the presence of a sulfur-containing compound
(thiophene) and hydrogen. Typically, ferrocene-assisted CVD produced MWNTs at
temperatures around 800C and a mixture of SWNTs and MWNTs at temperatures
above about 1000C. The thiophene increased the yield of SWNTs; the temperature
and ow rates were optimized to produce long SWNT strands in a continuous process.
23
The strands contained roughly 5 wt% iron and amorphous carbon impurities. These
strands were tweezed into smaller ones 5 to 20 m in diameter. SEM showed a
structural hierarchy in the bers, which consisted of thousands of well-aligned
bundles of SWNTs (diameter 1.1 to 1.7 nm) arranged in a two-dimensional trian-
gular lattice. The volume fraction of SWNTs in the strands was estimated to be less
than 48% based on the spacing between ropes measured by SEM. Raman spectroscopy
showed that the dominant SWNT diameter was 1.1 nm.
23
X-ray diffraction showed a
full-width half-maximum (FWHM) of 44.
24
Electrical resistivity on strands ranging
in diameter from 50 m to 0.5 mm was measured with a four-probe method. The
crossover temperature (metallic to semiconducting) was approximately 90 K; from 90
to 300 K the behavior was metallic with resistivity = 0.55 to 0.7 m cm.
23
Tensile
testing was complicated by difculties in measuring the cross-section of the strand
(and thus the true stress) during testing. The starting diameter was measured by SEM
and was in the range of 520 m. The estimated Youngs modulus ranged from 49 to
77 GPa. The authors assert that if one bases the Youngs modulus on a cross-section
corresponding to a nanotube volume fraction of less than 48%, the result is 150 GPa,
consistent with the modulus of SWNT bundles.
23
Li et al. have demonstrated the ability to produce continuously SWNT and
MWNT bers during chemical vapor deposition.
25
This process entails continuously
winding an aerogel of high-purity nanotubes formed in the furnace hot zone onto a
rotating rod. By changing the angle of the rod relative to the furnace axis, bers
with different degrees of twist can be produced. As with any chemical vapor depo-
sition process, the reactor conditions can be controlled to produce either SWNTs or
MWNTs; thus either SWNT or MWNT bers can be produced by varying the reactor
conditions. However, the SWNT bers contain approximately 50 vol% impurities
compared to approximately 10% for the MWNT bers. Both types of bers showed
high degrees of alignment. Electrical resistivities as low as 0.12 m cm were
achieved. Mechanical strength varied from 0.1 to 1 GPa, assuming a typical carbon
ber density of 2 g/cm
3
.
25
11.3 ELECTROPHORETIC SPINNING
Gommans et al. have spun bers electrophoretically from puried laser vaporization-
grown SWNTs dispersed in n,n-dimethylformamide (DMF) at concentrations of
about 0.01 mg/ml.
26
A commercially available carbon ber (8 m in diameter, about
12 mm long) was connected to a conducting wire (0.5 mm diameter, 10 mm long)
with silver conducting paint. The wire was attached to a motor-driven stage via a
pin vise. The carbon ber translated along its axis into the suspension at a depth
of a few millimeters. The carbon ber was turned into a positive electrode by
applying voltage, causing the SWNTs to migrate toward it and to form a cloud
around the carbon ber. SWNTs migrate because they are negatively charged in
DMF and move electrophoretically toward the positively charged carbon ber. As
the carbon ber was slowly withdrawn from the suspension, another ber, attached
to its end, spontaneously formed from the SWNT cloud. The ber length was
limited by the travel distance of the translation stage, the size of the SWNT cloud,
and the smoothness of the withdrawal from the solution. Fibers were typically
several centimeters long with diameters between 2 and 10 m. The mass of SWNTs
below the bath surface and the surface tension in the meniscus promoted the
coalescence and axial alignment of bundles of SWNTs. Alignment was measured
by polarized Raman spectroscopy and the axial-versus-perpendicular ratio was in
the range of 2 to 6.
26
11.4 CONVENTIONAL FIBER SPINNING
There are three main types of commercial ber spinning processes (Fig. 11.1): melt
spinning, dry solution spinning, and wet solution spinning.
27
Melt spinning is used
to produce many polymer bers, such as nylon and polyethylene terephthalate bers,
for commodity applications such as residential carpeting; it can be applied to single-
component as well as composite bers. In melt spinning, the ber-forming material
is melted and extruded under tension, typically into cooled air; the rapid cooling
induces the solidication of the ber. Some polymer/SWNT composite bers of
which polymer is the major constituent have been produced using this method.
However, SWNTs decompose without melting at approximately 750C in air and
at approximately 2000C in an inert atmosphere (e.g., argon); thus melt spinning is
not a viable option for bers in which SWNTs are the sole components.
Most SWNT bers have been produced by the solution spinning process. Solution
spinning is more complicated than melt spinning because the solidication of the ber
involves additional steps. The ber-forming material must be dissolved or nely
dispersed into a solvent, and the solvent must be extracted after the extrusion to form
the solid ber. Therefore, solution spinning is typically used to produce bers from
materials that decompose before reaching their melting point or do not have a suitable
viscosity for stable ber formation. Solution spinning has been particularly effective
for spinning bers from stiff polymer molecules, e.g., poly(p-phenylene terephthala-
mide) (PPTA), that form liquid crystals when dispersed in a solvent.
Several research groups have used solution spinning to produce bulk quantities
of SWNT bers (continuous lengths of 1 m or longer). Solution spinning can be
considered as a four-step process: (1) dispersion or dissolution of the ber material
into a solvent, (2) mixing and spinning the dispersion, (3) coagulation and drawing
into a solid ber, and (4) postprocessing of the ber through subsequent washing,
drying, or annealing steps.
Solution spinning can be divided into two main categories: dry solution spinning
and wet solution spinning (Fig. 11.1). In both cases, the spinning solution consists
of the polymer dissolved in a solvent. The solution is extruded through one or more
small orices in the spinneret and the solvent is removed, solidifying the ber.
Typically the polymer content is less than 30 wt% and sometimes less than 10 wt%.
In many cases, the solution is formed at the molecular level. Dry spinning is used
for systems such as cellulose acetate in acetone where the solvent is sufciently
volatile that it can be evaporated rapidly from the ber during its formation. Wet
spinning is used when the polymer is dissolved in a nonvolatile solvent that must
be extracted by using another liquid that is miscible with the solvent but cannot
dissolve the ber-forming material. A key example of a wet solution-spun ber is
DuPont Kevlar (Wilmington, DE), produced from poly(p-phenylene terephthala-
mide) (PPTA) dissolved in sulfuric acid. In wet solution spinning (Fig. 11.1), the
solution typically goes through an air (or inert gas) gap and then enters the coagulation
bath. In this case, the process is termed dry-jet wet spinning. The air gap allows
elongation and cooling of the spinning solution prior to coagulation. The presence
and length of the air gap depend on the ability of the polymer solution exiting the
FIGURE 11.1 Three main types of spinning processes: melt spinning, dry solution spin-
ning, and wet solution spinning.
27
(Reprinted with permission from Griskey, R.G., Polymer
Process Engineering, Chapman & Hall, New York, 1995. Copyright CRC Press, Boca Raton,
Florida.)
Feed
Melting
Metered
extrusion
Metered
extrusion
Metered
extrusion
Cooling and
solidifying
Moisture
conditioning
Lubrication
Yarn driving
Packaging
Packaging
Polymer
chips
Heating
grid
Pool
Pump
Filter and
spinneret
Filter and
spinneret
Air
diffuser
Steam
chamber
Heated
chamber
Oil disc
and trough
Feed
rolls
Bobbin
Bobbin
drive roll
(a) Melt spinning (b) Dry spinning (c) Wet spinning
Feed Feed
Solidification
by solvent
evaporation
Solidification;
decomposition
of soluble
deivatives;
or solvent
removal
Lubrication
Yarn driving
Balloon guide
Ring and traveler
Bobbin traverse
Spindle
Filtered
polymer
solution
Pump
Feed roll
and guide
Air
intake
Stretch
Packaging
Wet treating
Washing
Drying
Filtered polymer
derivative or
polymer solution
Pump
Filter and spinneret
Coagulation or
regeneration bath
Batch process
Bobbin
or cake
Forced circulation of
reagents, or water
Cakes
Drying
oven
spinneret to withstand surface forces prior to entering the coagulation bath. Liquid
laments are unstable to surface waves, which are amplied by surface tension
(Rayleigh instability);
28
higher surface tension and small lament diameters promote the
growth of the waves and the eventual breakup of the lament into droplets, whereas
viscous forces delay the development of the instability. High-viscosity bers as well
as thicker bers can normally support a larger air gap, particularly if the polymer is
extension-thickening, i.e., its viscosity in extension grows with strain rate. Typical air
gaps used in PPTA spinning are less than 2 cm; a gap of 0.5 to 1.5 cm is considered
optimal.
29
If larger gaps are used, the (liquid) ber breaks before entering the coagu-
lation bath. In contrast, higher viscosity poly(p-phenylene benzobisthiazole)
(PBZT)/poly(phosphoric acid) and (p-phenylene benzobisoxazole) (PBO)/poly(phos-
phoric acid) spinning solutions can support air gaps of up to 20 cm without ber line
breakage.
29
If the ber cannot support an air gap at all, the spinneret is immersed
directly into the coagulation bath and the process is termed wet-jet wet spinning.
Upon entering the coagulation bath, mass transfer occurs between the coagulant
(nonsolvent) in the bath and the solvent in the spinning solution, which can now be
considered the protober.
30,31
The goal of the coagulation process is to freeze the
structure of the protober and solidify the ber. The nal properties of the resulting
ber depend on the initial polymer solution, the coagulation process,
29,3235
drawing
of the ber, and postprocessing steps such as secondary washing
29,34
or drying
36
of
the ber.
The rst step of the solution spinning process, dispersal into a solvent, is far
from trivial for SWNTs.
7
Due to high van der Waals interactions, it is typically only
possible to disperse less than 0.5 wt% of SWNT in small molecule organic solvents
such as dichlorobenzene or aqueous solutions of surfactants such as sodium dodecyl
sulfate (SDS), although recent work has shown that superacids can disperse SWNTs
at higher concentration.
3739
Current research on SWNT functionalization and wrap-
ping may provide new means for dissolving higher concentrations of SWNTs in
these liquids for ber production.
4043
Finding an appropriate dispersion for the SWNTs is not sufcient for producing
high-quality bers; understanding the phase behavior of the dispersion is critical.
15,37,44,45
The phase behavior and the related rheological properties of the dispersion pro-
foundly inuence the selection of the spinning solution concentration, spinning and
coagulation process variables, and alignment in the as-spun bers. Ultimate ber
properties are determined by these variables as well as subsequent washing, drawing,
drying, and annealing processes.
46
11.4.1 MELT-SPUN COMPOSITE FIBERS
Reinforcement of existing materials with nanomaterials including SWNTs is a
topic of signicant current interest because relatively low loadings of the costly
component (SWNTs) could yield signicant improvements in properties. Improve-
ments in electrical and thermal conductivity are of particular interest, although
mechanical reinforcement is also an important application. As with the develop-
ment of other SWNT materials, this eld is still in its infancy; current results
should be viewed as preliminary and not indicative of the full potential of SWNT
composite bers.
Some of the rst melt-spun composite SWNT bers were produced by Andrews
et al. using carbon pitch as the main component.
47
SWNTs from Carbolex, Inc.
(Lexington, KY) were puried and mixed with isotropic petroleum pitch in quino-
line. After sonication and solvent removal, the bers were spun, drawn, and put
through a two-stage heat treatment in air and nitrogen. Mixtures containing 8 and
10 wt% SWNT yielded bers that were too weak to be collected; however, bers
produced with 1 wt% and 5 wt% SWNTs showed improved mechanical and electrical
properties. Improvements were a 90% increase in tensile strength, a 150% increase
in elastic modulus, and a 70% drop in electrical resistivity (Table 11.1).
Haggenmueller et al. produced SWNT polymethylmethacrylate (PMMA) lms
and bers.
48
Three types of SWNT were investigated: raw and puried soot from
laser ablation, and Tubes@Rice material. The SWNTs were dispersed in PMMA
(M
w
= 10,000) in dimethyl formamide (DMF) by prolonged sonication. After evap-
oration of the DMF, the resulting SWNT/PMMA pellets were melt mixed, spun
from a 600-m orice, and drawn under tension by a windup spool at high speed.
The maximum draw ratio (dened as = (D
o
/D)
2
) ranged from 60 to 3600, corre-
sponding to ber diameters between 75 and 10 m, respectively. Increasing SWNT
content dramatically increased the melt viscosity; this resulted in melt fracture
evidenced by surface roughness, striations along the ber axis, and nonuniform
diameter. Increasing the SWNT weight fraction also decreased the attainable draw
ratios and increased the frequency of ber breaks; bers were successfully produced
from loadings up to 8 wt% SWNT. The 10 wt% SWNT mixture was too viscous to
be extruded. Mechanical and electrical properties depended on SWNT concentration
and draw ratio. For melt-pressed lms, at a draw ratio of = 4, the electrical
resistivity in the direction parallel to processing decreased approximately 100-fold
from 850 cm at 1.3 wt% SWNT to 8.7 cm at 6.6 wt%. In the direction perpen-
dicular to processing, the electrical resistivities were only slightly higher: 1280 and
14 cm, respectively. The bers were highly anisotropic; their conductivity was
below the detection limit, in agreement with the fact that weight fractions of SWNTs
for percolation increase as the nanotubes align. At a draw ratio of = 70, the elastic
modulus G increased from 3.1 GPa in the absence of SWNT to 6 GPa for 8 wt%
SWNT. The draw ratio had a small effect on the modulus but signicantly affected
the yield stress. In bers containing 5 wt% puried SWNT soot, the yield stress
doubled from 65 to 130 MPa as the draw ratio was raised from = 40 to 300.
Similar results were observed with the other types of SWNTs. These improvements
are consistent with the high degree of alignment measured by polarized Raman
TABLE 11.1
Properties of SWNT/Pitch Fibers
45,47
Sample
Tensile Strength
[MPa]
Elastic Modulus
[GPa]
Electrical Resistivity
[m cm]
Pitch 480 33 8.4
Pitch/1 wt% SWNTs 620 41 2.5
Pitch/5 wt% SWNTs 850 77 2.2
spectroscopy at multiple angles relative to the ber axis; a ber containing 1 wt%
puried soot had a full width half maximum (FWHM) of 4.
48
Efforts to incorporate SWNTs into polypropylene (PP) bers have built on earlier
success in incorporating carbon nanobers into thermoplastics.
4952
Bhattacharyya
et al.
53
produced 0.8 wt% SWNT-PP bers using HiPco SWNTs and polypropylene
of 17 melt ow index (MFI). The SWNTs were puried (less than 1 wt% impurities)
using methods found in the literature,
54
mixed with the PP in a Haake Rheomix 600
at 240C, and ltered to remove the largest aggregates. This preparation procedure,
however, failed to disperse satisfactorily the SWNTs that remained clumped into
domains of tens of micrometers. The SWNTs were found to promote faster PP crys-
tallization, probably by acting as nucleation sites, and to yield smaller PP crystallites.
The SWNTs in the nal ber were better oriented axially than the PP. Mechanical
properties were largely unaffected by the presence of the SWNTs as a result of the
poor dispersion. The mechanical properties are expected to improve as better methods
for dispersing the SWNTs in PP are found.
11.4.2 SOLUTION-SPUN SWNT FIBERS
11.4.2.1 SWNT/Liquid Crystalline Polymer Composite Fibers
Improving the properties of high-performance bers such as poly(p-phenylene
benzobisthiazole) (PBO) through the incorporation of SWNTs is an important
potential application of SWNTs. Commercial PBO ber (Zylon) is one of the
strongest bers with a tensile strength of 5.6 GPa.
55
Kumar et al.
56
synthesized
PBO in poly(phosphoric acid) (PPA) in the presence of puried HiPco SWNTs
at concentrations of 0, 5, and 10 wt%. Optical microscopy showed that the SWNTs
were well dispersed in the liquid crystalline PBO/PPA solution. The liquid crys-
talline solution was dry-jet wet spun into a water coagulation gap with an air gap
of 10 cm and draw ratios as high as 10. SWNT loading improved modulus, tensile
strength, and elongation to break. For 10 wt% the improvements were 20, 60,
and 40%, respectively (Table 11.2). Achieving similar improvements with com-
mercial Zylon bers (twice as strong as the laboratory control) could yield bers
with a tensile strength in excess of 8 GPa. Very importantly the incorporation of
SWNTs improved the compressive strength of the bers. The coefcient of
TABLE 11.2
Mechanical Properties of SWNT/PBO Fibers
56
Sample
Tensile
Modulus (GPa)
Strain to Failure
(%)
Tensile
Strength (GPa)
Compressive
Strength (GPa)
PBO 138 20 2.0 0.2 2.6 0.3 0.35 0.6
PBO/5 wt% SWNTs 156 20 2.3 0.3 3.2 0.3 0.40 0.6
PBO/10 wt% SWNTs 167 15 2.8 0.3 4.2 0.5 0.50 0.6
Reprinted with permission from Kumar, S., et al. Macromolecules, 35, 90399043, 2002.
Copyright 2002, American Chemical Society.
thermal expansion dropped from 6 ppm/C for the pure PBO ber to 4 ppm/C
at 10% SWNT loading. Thermal degradation was not affected by the presence
of SWNTs. The degree of alignment of the PBO in the ber was also unaffected
by the SWNTs. The electrical conductivity was below the limit of detection. The
authors state that such low conductivity indicates both good dispersion and strong
alignment; however, either one of these two morphological features would sufce
to explain the lack of conductivity, because 10% is well below the percolation
threshold of aligned slender anisotropic objects as well as randomly packed
isotropic ones.
57
Polyacrylonitrile (PAN) copolymers are used as carbon ber precursors and for
developing porous activated carbon that can be used in various applications (e.g.,
catalysis, electrochemistry, separations, and energy storage). Sreekumar et al.
58
pro-
duced PAN/SWNT bers by dry-jet wet solution spinning. Composite bers with
10 wt% SWNT had dramatically better properties than the pure PAN bers. The
tensile modulus nearly doubled from 7.9 GPa (pure PAN ber) to 14.2 GPa (5 wt%
SWNT ber) and 16.2 GPa (10 wt% SWNT ber). Tensile strength was not improved
as much. The pure PAN ber had a strength of 0.23 GPa and the 5% and 10% bers
had strengths of 0.36 GPa and 0.33 GPa, respectively. All bers showed similar
elongation at break (about 10%). The presence of SWNTs improved thermal stability.
The glass transition temperature grew from 103C (pure ber) to 114C (5 wt%
ber) to 143C (10 wt% SWNT). The SWNT/PAN bers retained their elastic
modulus better than the pure PAN bers at high temperature. (Interestingly the 5%
SWNT bers retained modulus better than the 10% bers.) At 200C, thermal
shrinkage of the 10% ber was nearly half that of the control. Supercapacitor lms
of SWNT/PAN were produced by a similar dispersion technique followed by lm
casting in vacuum.
59
11.4.2.2 Fibers Produced from SWNT/Surfactant Dispersions
The rst truly scalable continuous process for making bers composed primarily or
solely of SWNTs was developed by a team at the Centre National de la Recherche
Scientique (CNRS) and the Universit Bordeaux.
60,61
This process produced bers
composed of a network of SWNTs and a polymer, typically polyvinyl alcohol (PVA).
The SWNTs were dispersed in an aqueous solution with the aid of surfactants and
then injected into a owing polymer (PVA) solution. The polymer can be removed
by washing the ber repeatedly and by heating the ber at temperatures between
the decomposition temperature of the polymer and that of the SWNTs. The original
process developed at CNRS employs a rotating polymer solution bath mounted on
a turntable;
6062
a modied process developed by the University of Texas at Dallas
uses a coowing stream of polymer solution.
63,64
SWNT suspensions are analogous to solutions of rod-like polymers or anisotropic
colloids.
44
Just as the nature of surfactants plays a critical role in the phase behavior
of classical colloids, the same is true for nanotubes. In the CNRS process, SWNTs
were sonicated in aqueous solutions of sodium dodecyl sulfate (SDS), an anionic
surfactant that adsorbs at the surfaces of SWNT bundles. Similar adsorption behavior
has also been observed with cationic surfactants such as tetratrimethylammonium
bromide. Vigolo et al.
44,61
studied the phase behavior of SWNTs produced by electric
arc, laser vaporization, and HiPco in SDS in order to identify the ranges of surfactant
and SWNT concentration that yielded dispersions with optimal processability. The
stabilization mechanism seems to be a balance between repulsive electrostatic forces
induced by the charged groups on the surfactants and attractive forces induced by
van der Waals forces between SWNTs (or bundles of SWNTs) and by micellar
depletion.
At low surfactant concentration, the number of surfactant molecules adsorbed
on the SWNTs present in small bundles after sonication was not sufcient to
counterbalance the long-range van der Waals forces van der Waals forces between
parallel cylinders decay with r
1.5
, where r is the separation between the cylinders.
65
Therefore, the bundles stick back together after sonication and form large dense
clusters. At high surfactant concentration, the excess surfactant that is not adsorbed
on the SWNT bundles forms nanometer-sized micelles. As two SWNT bundles
approach each other, they exclude micelles; this produces a difference of osmotic
pressure,
65,66
which in effect brings the bundles closer together and causes the
formation of ocs that coexist with well-dispersed thin bundles.
44
The proportion of
clusters grows with surfactant concentration.
At intermediate surfactant concentration, the system forms an apparently homo-
geneous phase of well-dispersed small bundles. In the case of electric arc-produced
SWNTs, this homogeneous phase exists up to approximately 0.35 wt% SWNT in
1 wt% SDS. Similar behavior was observed for SWNTs produced by all three
production methods, but the exact phase boundaries were dependent on the production
technique; the boundaries shifted to higher concentrations for the HiPco-produced
SWNTs, probably because of differences in length and diameter distributions. The
homogeneous dispersions with the highest content of SWNTs showed optimal pro-
cessability, although bers of HiPco SWNTs could be produced also from nonho-
mogeneous dispersions.
44
The origin of the SWNTs also inuenced the ribbon/ber
morphology. For HiPco-produced tubes, a smaller weight fraction (approximately
0.15 wt%) was needed to make homogeneous bers with a uniform and nearly
cylindrical cross-section.
62
HiPco tubes also yielded stable ber formation and good
mechanical properties at higher velocities.
60
Due to their anisotropy, SWNTs align in shear or elongational ow. Injecting
the SWNT dispersion in a coowing stream of polymer has two benecial effects
on the ber formation. It aligns the SWNTs by imposing a small amount of exten-
sional ow on the protober that can be controlled by the mismatch between the
velocity of the SWNT dispersion being injected and the velocity of the polymeric
bath. Moreover the rotating bath supports and transports the protober during the
early stages of coagulation, while it is still too weak to support an applied tension.
If the coagulation bath were not moving, the ber would have to be drawn mechan-
ically through the coagulation bath, which is not possible unless the liquid lament
has a high enough viscosity and the ability to withstand tension along the streamlines
(usually associated with extensional thickening). Most of the surfactant leaves the
lament during the coagulation, while at the same time the PVA molecules pene-
trate the forming ber and stabilize it by forming bridges with the SWNTs.
61,62
The SWNT/SDS dispersions were injected through a 0.5-mm diameter syringe needle
into a rotating bath containing 2 to 6 wt% of PVA (M
w
= 70,000, hydrolysis 89%,
solution viscosity about 200 cP).
44,60,61
Amphiphilic polymers such as PVA are
ineffective at stabilizing SWNTs; they stick to each other almost as soon as they
enter the bath. This can be considered as coagulation through bridging occulation.
62
Polarized optical microscopy of ribbons suspended in water showed that there is
preferential alignment along the ribbon axis. SEM showed preferential axial align-
ment in the cleaned dry ribbons although the dried ribbons displayed weaker anisot-
ropy than the ribbons suspended in water. Low-viscosity PVA solutions yielded
bers with lower alignment, possibly because they were less effective at rapidly
stretching the extruded lament.
High strain rates in the processing ow yielded better aligned as-spun bers.
60
The relevant quantities include the shear rate in the syringe tube, the ratio of the
viscosity of the coagulant and SWNT dispersion, and the ratio between the coagulant
velocity and the extrusion speed of the SWNT dispersion. The shear rate in the
syringe tube should promote axial alignment of the SWNT bundles before they meet
the PVA solution. A higher coagulant/SWNT dispersion viscosity ratio should accel-
erate the lament over a smaller distance, i.e., produce a higher rate of extension
immediately after the syringe orice. The ratio of the velocity of the rotating PVA
bath and the extruded SWNT dispersion controls the total extensional strain that is
experienced by the lament as it coagulates into a ribbon.
60
However, variations in
polymer solution speed between 300 and 800 m/h and the SWNT dispersion injection
rate (35100 ml/hr) did not affect the orientation in the dried bers. This suggests
that alignment is xed during the drying stage and not during the initial coagulation
process.
62
The bers from the CNRS process possess a hierarchical structure with well-
developed porosity and high surface area (160 m
2
/g by the Brunauer-Emmett-Teller
(BET) method, compared to 200 m
2
/g for standard bucky paper).
67
This makes them
of interest for nanouidic devices, adsorbents and absorbents, catalysis support, and
host systems for encapsulation of biomolecules for sensors, smart textiles, and other
nanotechnologies in which wetting and sorption properties play an important role.
Cross-sections of bers dried under tension reveal a well-dened core consisting
mostly of SWNTs and a shell that contains most of the carbon impurities; such
impurities were randomly distributed in the initial bers. Thus, separation occurred
as the bers collapsed and this may provide a means of large-scale SWNT puri-
cation. Chemical irradiative or thermal removal of the shell would result in puried
SWNTs.
61
Neimark et al.
67,68
made detailed studies of the ber morphology and gas adsorp-
tion. Gas adsorption measurements were made on bers that had been heated in air
at 320C for 3 h. This removed 95% of the organic species while keeping the bers
intact. The bers have several levels of structural organization; 10- to 50-m bers
are composed of straight elementary laments 0.2 to 2 m in diameter with a hairy
surface.
67
These elementary laments consist of densely aligned nanotube bundles
1030 nm in diameter, which in turn consist of SWNTs 12 nm in diameter. The
elementary laments are aligned in a closely packed conguration in the ber skin
of 15 m diameter, while the core consists of a nanofelt of loosely packed bundles;
the skin-core morphology resembles that of polyacrylonitrile (PAN)-based carbon
bers. This was conrmed by comparing alignment measured by x-ray scattering and
polarized Raman spectroscopy. In x-ray scattering, where the whole ber is exposed
to radiation, the FWHM was 75. In polarized Raman spectroscopy, where a limited
boundary layer of about 1 m is probed, the FWHM was about 35; this indicates
that the surface alignment is much better than the alignment in the core.
The original CNRS process resulted in bers that showed a plastic deformation
before breaking and could be curved through 360 within a length of tens of
micrometers, demonstrating that the bers are more exible and resistant to torsion
than classical carbon bers.
61
Four-probe electrical resistivity measurements gave
a room temperature resistivity of approximately 0.1 cm and a nonmetallic behav-
ior when the temperature was decreased. This value was inuenced by the presence
of the polymer and carbon impurities; it was three orders of magnitude higher than
those of small bundles of SWNTs.
61
In an improvement to the original process, the
washed and dried bers were rewetted, swollen, and dried under tensile load with
a weight attached to one end of the ber. They could be stretched up to 160% and
did not disassemble even in solvents where the polymer was highly soluble. This
suggests that the SWNTs and adsorbed polymer form an effectively cross-linked
network that can be easily deformed.
62
The qualitative effect of stretching on ber
morphology and alignment can be seen in Fig. 11.2. This effect can be quantied
by the change in FWHM from 75 to 80 for raw bers to less than 50 for bers
with draw ratios of 125% or greater. At higher draw ratios, the SWNTs seemed to
slide and unbind from the PVA; both the applied stress needed to achieve the higher
draw ratio and the FWHM remained constant. The improvement in alignment with
stretching translated into improved mechanical properties: The Youngs modulus
increased four-fold from about 10 GPa for the raw bers to 40 GPa for bers
stretched 145% in water with a load of 0.44 g. The tensile strength of these bers
increased from 125 MPa to 230 MPa, respectively. The effectiveness of the stretching
process was found to be dependent on the liquid used to rewet the ber; poorer PVA
solvents yielded better alignment. For a draw ratio of 117%, increasing the concen-
tration of acetone (poor PVA solvent) in water (good PVA solvent) from 0% to 70%
decreased the FWHM from 65 to 50. Vigolo et al. hypothesize that in a poorer
solvent there is less sliding and the cohesion of the SWNT-polymer network is
better.
62
The improved CNRS process has also been used to spin bers from DNA-stabilized
SWNT dispersions.
69
Typical ber dimensions are 50 cm in length and 20 to 30 m in
diameter. Youngs modulus as high as 19 GPa, and tensile strength as high as 130 MPa
were achieved. These values did not vary signicantly with stretching; they are two to
three times better than those the authors achieved for unstretched bers spun from
SWNT-SDS dispersions. However, these values are similiar to to those achieved by
Vigolo et al. (2002) for unstretched SWNT-SDS bers and lower than those Vigolo et
al. achieved for stretched SWNT-SDS bers. The electrical resistivity of bers produced
from the DNA-stabilized dispersions was more than 50% greater than resistivities of
the bers produced from the SDS-stabilized suspensions.
69
Baughman and coworkers at the University of Texas at Dallas have built on the
CNRS work and developed a process for producing SWNT bers, primarily from HiPco
SWNTs dispersed in aqueous lithium dodecyl sulfate with PVA as a coagulant.
7,63,64
In this process, the spinning dispersion is injected into a stream of coagulant that is
owing in a pipe coaxial with the spinning needle. The gel protobers are wound onto
a mandrel and then passed through a series of acetone and water baths for drawing and
solvent removal. The gel protobers can be continuously processed at approximately
100 cm/min to produce reels of several hundred continuous meters of 50-m diameter
bers consisting of approximately 60 wt% SWNT and 40% PVA.
70
Like Vigolo et al.,
Baughman et al. also found that HiPco tubes spin over a much broader concentration
range and produced more mechanically robust ribbons.
63
The bers have impressive mechanical and electrical actuation properties.
64,70
The PVA forms a coating on the SWNTs and the mechanical properties are believed
to result from internanotube stress transfer via the PVA.
70
Tensile strengths up to
3.2 GPa have been achieved; this is nearly twice that of spider silk, and about ten
times higher than that of the CNRS bers. The Youngs modulus is 80 GPa, which
equals that of the bers grown by Zhu et al.
23
and is double that of bers spun by
Vigolo et al.;
62
however, this value is still roughly an order of magnitude less than
that of high-performance graphite bers or individual SWNTs. Normalized for
density, the Youngs modulus and tensile strength of these bers are more than twice
FIGURE 11.2 Effect of drying under load on alignment. (Reprinted with permission from
Vigolo, B. et al. Appl. Phys. Lett., 81, 12101212, 2002. Copyright American Institute of
Physics, Melville, NY.)
that of steel and the bers are approximately 20 times as tough as steel.
64,70
Fiber
toughness, the total mechanical energy a specied mass of ber absorbs before
breaking, surpasses that of spider silk, the benchmark for ber toughness. The
predrawn SWNT/PVA bers match the energy absorption of spider silk with a 30%
breaking strain and continue to absorb energy until reaching an energy to break of
over 600 J/g, which is higher than the best spider drag line silk (165 J/g).
64
The high
elongation required for toughness is achieved because the ber does not neck during
loading; testing showed that strain rate was linearly dependent on stress.
64
Dalton et al.
64
anticipate extending their technology to coagulation polymers other
than PVA, particularly those with higher modulus or ionic conductivity. Initial com-
mercial applications are anticipated in niche (low volume/high value) applications such
as actuators for microcatheters for microsurgery, moving pins on dynamic braille
displays, three-dimensional optical ber switches, and as minority components for
electronic textiles.
64
The ability to make the bers into supercapacitors and weave them
into textiles has been demonstrated; 100-micron supercapacitors provide high life cycle
and comparable energy storage to large aqueous-electrolyte supercapacitors. However,
the slower discharge rates indicate that higher ber electrical conductivity must be
achieved.
70
Other potential applications for weavable and sewn SWNT/PVA bers are
distributed sensors, electronic interconnects, electromagnetic shields, antennas, and
batteries.
64
Fiber toughness could be exploited in antimeteorite/antiballistic shields for
satellites, antiballistic vests, and explosion-proof blankets for aircraft cargo bays.
Longer term applications include actuators (articial muscles) for prosthetics and severe
environments such as those encountered in planetary exploration and aircraft engines.
70
11.4.2.3 Fibers Produced from SWNT/Superacid Dispersions
Highly aligned bers consisting solely of SWNTs have been produced by wet spinning
SWNT/superacid dispersions.
37,45,71,72
The direct protonation of single-walled carbon
nanotubes in superacids allows them to be dispersed at more than an order of magnitude
higher concentration than that typically achieved in surfactants or organic sol-
vents.
3739,72
Rheology and microscopy show that SWNTs in superacids roughly par-
allel the phase behavior of rod-like polymer solutions (such as PPTA in sulfuric acid)
used for solution spinning of high-performance bers from nematic liquid crystals.
37
The phase behavior of rod-like polymer solutions is depicted in Fig. 11.3. With
increasing concentration, such a system transitions from a dilute solution where
FIGURE 11.3 Phase behavior of rigid rods in solution.
37
Reprinted with permission from
Macromolecules, 37, 154160, 2004. Copyright 2004 America Chemical Society.
Increasing concentration,
Dilute Semi-
dilute
Liquid
cryst.
Isotropic
conc.
Biphasic
individual rods do not interact with each other, to a semidilute solution where rotation
is inhibited, to an isotropic concentrated solution where both translation and rotation
are inhibited. At concentrations above the percolation threshold, steric effects cause
the system to phase separately into a liquid crystalline phase in equilibrium with the
isotropic concentrated phase.
73,74
Raising the concentration further increases the pro-
portion of the liquid crystalline phase until the system becomes fully liquid crystalline.
In superacids, SWNTs form charge-transfer complexes of positively charged nano-
tubes surrounded by acid anions.
38
Such charge-transfer complexes repel each other
electrostatically at short distance; however, the charges are screened by the acid; thus,
electrostatic repulsion is expected to decay rapidly with SWNT separation, and the
SWNTs can attract each other at longer separation distances through van der Waals
forces (which have long range in the case of long, nite-diameter cylinders
65
such as
SWNTs). Such long-range forces, together with excluded volume, seem responsible
75
for the inception of the biphasic region at much lower concentration than would be
predicted by the theory for noninteracting rods.
37
The biphasic region consists of aligned,
mobile, thread-like SWNT/acid charge-transfer complexes surrounded by sulfuric acid
anions in equilibrium with an isotropic phase. The aligned phase has been dubbed
SWNT spaghetti. At high enough concentration, the spaghetti form a single-phase
nematic liquid crystal; this occurs at approximately 4 wt% SWNT for puried
54
HiPco
SWNTs with an average aspect ratio L/D of 470 dispersed in 102% sulfuric acid.
This liquid crystal has a birefringent polydomain structure and key rheological signa-
tures such as a maximum in the viscosity versus concentration curve, long oscillatory
transients, and the rst normal stress difference changing sign from positive to negative
and back to positive with increasing shear rate.
37
Moreover, a variety of aligned species
can be formed from the dispersion. Liquid crystals of oxidized MWNTs
76
and SWNTs
in isopropyl acrylamide gel
77
have also been reported.
Production of SWNT bers from 102% H
2
SO
4
requires mixing of the dispersion
in an anhydrous environment. The introduction of moisture destabilizes the dispersion
resulting in the formation of tactoid-shaped aggregates dubbed alewives, roughly
50 m long and 2 m wide.
37,38,78
The alignment in the dispersion facilitates the
production of highly aligned neat SWNT bers even in the absence of drawing. Simply
pushing the dispersion through a 125-m syringe needle into a diethyl ether coagulation
bath results in the formation of bers with 12: 1 axial to radial alignment as measured
by Raman spectroscopy.
45
The development of an integrated mixing and spinning
apparatus for producing SWNT bers from SWNT acid dispersions has resulted in
bers with better alignment and properties.
45,72
Fibers have been produced from 4 to
10 wt% puried SWNT in 102% sulfuric acid. The maximum achievable SWNT
loading has been dictated by the ability to mix and extrude the dispersion. The viscosity
of the 6 wt% dispersion is roughly 1000 Pa s at 0.1 s
1
(one million times that of water)
at room temperature, and heating has little effect on the viscosity.
Typical coagulants for the SWNT/superacid dispersions are diethyl ether, dilute
sulfuric acid solutions, and water. Diethyl ether is an effective coagulant for these bers
because it is miscible with sulfuric acid and it evaporates rapidly in air upon removal
from the bath. During coagulation, the SWNTs do not have time to reorient and the
ber quickly collapses into a dense structure.
45
Both dilute sulfuric acid solutions and
water facilitate mass transfer of the 102% sulfuric out of the ber and result in a denser
ber of more uniform cross section. Water-coagulated bers spun from 8 wt% SWNT
in 102% H
2
SO
4
had a density of 1.1 g/cc, approximately 77% of the theoretical close-
packed density for 1.0-nm SWNTs (1.5 g/cc).
72
In all cases, annealing at high temper-
atures (on the order of 1100C) is needed to remove residual sulfur.
13,72
Scanning electron microscopy (SEM) of bers produced from 610 wt% SWNT
show signicant alignment and a consistent, clean surface morphology generally
free of impurities.
37,45
High-magnication images of the neat SWNT bers reveal a
ber substructure of aligned SWNT bundles approximately 200600 nm in diameter
and a bundle substructure of ropes on the order of 20 nm.
45,72
The bers possess a
relatively uniform circular cross section, which indicates a homogeneous coagulation
without brittle skin formation or structural collapse. Peeling of the bers as well as
cross-sectioning of the bers by microtome and UV-ozone etching
45,72,79
show that
the alignment and packing are consistent throughout the ber. This differs from the
SWNT/PVA bers where the surface is more aligned and has a different morphology
from the interior.
67
Fiber alignment as high as 28: 1 axial to perpendicular has been measured by
polarized Raman spectroscopy; the bers are the most highly aligned macroscopic
neat SWNT material produced to date.
45,80
This alignment is achieved without any
extensional drawing during ber production; it is largely the result of the liquid
crystalline nature of the SWNT/superacid dispersion. Zhou et al. determined that for
6 to 8 wt% puried SWNTs dispersed in 100% H
2
SO
4
and spun into diethyl ether,
decreasing extrusion orice diameter from 500 to 125 m improved alignment and
electrical and thermal transport.
13
The improvement in alignment was quantied by
the FWHM of wide-angle x-ray scattering peaks, decreasing from 63 to 45 and the
aligned fraction as measured by Raman increasing from 0.83 to 0.94.
13
Subsequent
improvements in ber spinning have resulted in even more highly aligned bers
(FWHM of 31).
72
Annealing the bers at 1100C in argon for 24 hours or at 1150C
in vacuum for 2 hours had little effect on the alignment. Annealing also had little to
no effect on the thermal conductivity of the bers. Thermal conductivities at 300 K
were 17 W/mK and 5 W/mK for the most and least aligned bers, respectively.
13
These
low values, compared to that of graphite parallel to the layers, are likely the result of
several factors. The cross-sectional area was based on the macroscopic dimensions
and did not account for gross voids or porosity. Also the bers are not perfectly aligned,
and the gaps between bundles represent thermal barriers.
13
Unlike alignment and thermal conductivity, electrical properties were signicantly
affected by annealing. The as-produced bers were heavily p-doped with low resis-
tivity and metallic behavior above 300 K; resistivity decreased from 0.64 to 0.25 m
cm at 300 K for the 500- and 125-m orices, respectively. Annealing the bers
removed the residual acid dopants causing an order of magnitude increase in resistivity
at 300 K and a nonmetallic temperature dependence.
13
Thermoelectric power (S)
measurements showed an unusual low-temperature behavior explained by a one-
dimensional phonon-drag model.
81
The two characteristic features of a phonon-drag
contribution are the upturn in the thermopower at T
0
0.1
Q
/k
B
, where
Q
is the
energy of the relevant phonon mode, and a peak around dS/dT 2.3T
0
. The bers
from SWNT/acid show an upturn in thermopower at T
0
= 26 K and a differential
temperature maximum at 4955 K. Modeling the data by a sum of phonon-drag
thermopower and diffusion thermopower resulted in the phonon energy
Q
= 0.02
eV, diffusion thermopower coefcient b = 0.04 V/K
2
, and L
ph
(Q) = 0.6 and 0.4 m
at 10 and 300 K, respectively.
Fiber mechanical properties depend on SWNT concentration, orice diameter
(alignment), and coagulant. A Youngs modulus of 120 GPa and tensile strengths of
120 MPa have been achieved in bers on the order of 50 m in diameter and tens
of meters in length.
72
The tensile strength is limited by the presence of defects, voids,
and rope-rope interfaces, and was improved by a factor of 30 over the course of one
year.
45
Further optimization of the dispersion, spinning conditions, coagulation con-
ditions, and postprocessing are expected to reduce ber defects and improve mechan-
ical properties.
11.5 CONCLUSION
The development of SWNT nanotube bers is in its infancy, and the performance
expected from the nanoscale properties of SWNTs has yet to be realized in macro-
scopic objects. However, preliminary composite and neat bers show promising
results. It is likely that spinning processes will evolve to meet specic applications
needs. SWNTs dispersed in a polymer matrix may be best suited for extending the
potential applications of existing polymeric materials by enhancing thermal and
electrical conductivity and compressive strength at a relatively low-cost differential.
SWNT/surfactant dispersions may be solution spun for applications where toughness
and other mechanical properties are most critical. SWNT/acid dispersions used to
produce neat bers may be best suited for applications with the most demanding
thermal and electrical requirements. The development of these spinning technologies
along with improvements in SWNT production, purication, and functionalization
will undoubtedly result in a number of revolutionary ber applications.
PROBLEMS
1. Describe a potential application for nanotube bers. What properties
would be required? What kind of nanotubes would you use? Which of
the ber production methods described in this section would you use? (At
this stage, assume that cost is not an issue.)
2. How does the polydispersity in chirality, length, and diameter of SWNTs
affect the different ber spinning processes? How does this polydispersity
affect ber properties?
3. What are the biggest challenges in producing nanotube polymer composite
bers? How do these differ from the challenges in producing neat SWNT
bers?
ACKNOWLEDGMENTS
For discussions and for sharing ideas and experimental results we thank Lars Ericson,
Sivarajan Ramesh, Rajesh Saini, Nicholas Parra-Vasquez, Hua Fan, Yuhuang
Wang, Haiqeng Peng Wei Zhou, Carter Kittrell, Bob Hauge, Wade Adams, Ed Billups,
Boris Yakobson, Howard Schmidt, Wen-Fang Hwang, Rick Smalley, Satish Kumar,
Karen Winey, Jack Fischer, Alan Dalton, Ray Baughman, Edgar Munoz, Philippe
Poulin, and Erik Hobbie. We acknowledge funding from the DURINT initiative of
the Ofce of Naval Research under Contract N00014-01-1-0789, from the Texas
Advanced Technology Program through grant 003604-0113-2003, and from the
National Science Foundation through Award EEC-0118007.
REFERENCES
1. Tans, S.J., Devoret, M.H., Dai, H.J., Thess, A., Smalley, R.E., Geerligs, L.J., and
Dekker, C., Individual single-wall carbon nanotubes as quantum wires, Nature, 386,
474477, 1997.
2. Hone, J., Llaguno, M.C., Nemes, N.M., Johnson, A.T., Fischer, J.E., Walters, D.A.,
Casavant, M.J., Schmidt, J., and Smalley, R.E., Electrical and thermal transport
properties of magnetically aligned single wall carbon nanotube lms, Appl. Phys.
Lett., 77, 666668, 2000.
3. Berber, S., Kwon, Y.K., and Tomanek, D., Unusually high thermal conductivity of
carbon nanotubes, Phys. Rev. Lett., 84, 46134616, 2000.
4. Hone, J., Whitney, M., Piskoti, C., and Zettl, A., Thermal conductivity of single-
walled carbon nanotubes, Phys. Rev. B, 59, R2514R2516, 1999.
5. Krishnan, A., Dujardin, E., Ebbesen, T.W., Yianilos, P.N., and Treacy, M.M.J., Youngs
modulus of single-walled nanotubes, Phys. Rev. B, 58, 1401314019, 1998.
6. Buongiorno-Nardelli, M., Yakobson, B.I., and Bernholc, J., Brittle and ductile behav-
ior in carbon nanotubes, Phys. Rev. Lett., 81, 46564659, 1998.
7. Baughman, R.H., Zakhidov, A.A., and de Heer, W.A., Carbon nanotubesthe route
toward applications, Science, 297, 787792, 2002.
8. Zhang, X.F., Cao, A.Y., Li, Y.H., Xu, C.L., Liang, J., Wu, D.H., and Wei, B.Q., Self-
organized arrays of carbon nanotube ropes, Chem. Phys. Lett., 351, 183188, 2002.
9. Salvetat-Delmotte, J.P. and Rubio, A., Mechanical properties of carbon nanotubes: a
ber digest for beginners, Carbon, 40, 17291734, 2002.
10. Bronikowski, M.J., Willis, P.A., Colbert, D.T., Smith, K.A., and Smalley, R.E., Gas-
phase production of carbon single-walled nanotubes from carbon monoxide via the
HiPco process: a parametric study, J. Vac. Sci. Technol. A Vac. Surf. Films, 19,
18001805, 2001.
11. Dai, H.J., Nanotube growth and characterization, Topics Appl. Phys. Carbon Nano-
tubes, 80, 2953, 2001.
12. Kitiyanan, B., Alvarez, W.E., Harwell, J.H., and Resasco, D.E., Controlled production
of single-wall carbon nanotubes by catalytic decomposition of CO on bimetallic Co-
Mo catalysts, Chem. Phys. Lett., 317, 497503, 2000.
13. Zhou, W., Vavro, J., Guthy, C., Winey, K.I., Fischer, J.E., Ericson, L.M., Ramesh,
S., Saini, R., Davis, V.A., Kittrell, C., Pasquali, M., Hauge, R.H., and Smalley,
R.E., Single wall carbon nanotube bers extruded from super-acid suspensions:
preferred orientation, electrical, and thermal transport, J. Appl. Phys., 95, 649655,
2004.
14. Yakobson, B.I. and Smalley, R.E., Fullerene nanotubes: C
1,000,000
and beyond, Am.
Scientist, 85, 324337, 1997.
15. Baughman, R.H., Materials scienceputting a new spin on carbon nanotubes, Science,
290, 13101311, 2000.
16. Baughman, R.H., Materials sciencemuscles made from metal, Science, 300,
268269, 2003.
17. De Zhang, W., Wen, Y., Liu, S.M., Tjiu, W.C., Xu, G.Q., and Gan, L.M., Synthesis
of vertically aligned carbon nanotubes on metal deposited quartz plates, Carbon, 40,
19811989, 2002.
18. Thostenson, E.T., Ren, Z.F., and Chou, T.W., Advances in the science and technology
of carbon nanotubes and their composites: a review, Compos. Sci. Technol., 61,
18991912, 2001.
19. Li, X.H., Zhang, J., Li, Q.W., Li, H.L., and Liu, Z.F., Polymerization of short single-
walled carbon nanotubes into large strands, Carbon, 41, 598601, 2003.
20. Li, Y.H., Wei, J.Q., Zhang, X.F., Xu, C.L., Wu, D.H., Lu, L., and Wei, B.Q., Mechan-
ical and electrical properties of carbon nanotube ribbons, Chem. Phys. Lett., 365,
95100, 2002.
21. Zhang, Y.G., Chang, A.L., Cao, J., Wang, Q., Kim, W., Li, Y.M., Morris, N., Yenilmez,
E., Kong, J., and Dai, H.J., Electric-eld-directed growth of aligned single-walled
carbon nanotubes, Appl. Phys. Lett., 79, 31553157, 2001.
22. Jiang, K.L., Li, Q.Q., and Fan, S.S., Nanotechnology: spinning continuous carbon
nanotube yarnscarbon nanotubes weave their way into a range of imaginative
macroscopic applications, Nature, 419, 801801, 2002.
23. Zhu, H.W., Xu, C.L., Wu, D.H., Wei, B.Q., Vajtai, R., and Ajayan, P.M., Direct
synthesis of long single-walled carbon nanotube strands, Science, 296, 884886,
2002.
24. Wei, B.Q., Vajtai, R., Choi, Y.Y., Ajayan, P.M., Zhu, H.W., Xu, C.L., and Wu, D.H.,
Structural characterizations of long single-walled carbon nanotube strands, Nano
Lett., 2, 11051107, 2002.
25. Li, Y.L., Kinloch, I.A., and Windle, A.H., Direct spinning of carbon nanotube bers
from chemical vapor deposition synthesis, Science, 304, 276278, 2004.
26. Gommans, H.H., Alldredge, J.W., Tashiro, H., Park, J., Magnuson, J., and Rinzler,
A.G., Fibers of aligned single-walled carbon nanotubes: polarized Raman spectros-
copy, J. Appl. Phys., 88, 25092514, 2000.
27. Griskey, R.G., Polymer Process Engineering, Chapman & Hall, New York, 1995,
chap. 11.
28. Levich, V.G., Physiochemical Hydrodynamics, Prentice Hall, Englewood Cliffs, 1962,
chap. 8.
29. Rakas, M.A., The Effect of Coagulants on the Microstructure and Mechanical Prop-
erties of Lyotropic Fiber-Forming Polymers, Ph.D. Thesis, Department of Polymer
Science and Polymer Engineering, University of Massachusetts, available through
UMI, 1990.
30. Paul, D.R., Diffusion during the coagulation step of wet-spinning, J. Appl. Polym.
Sci., 12, 383402, 1968.
31. Ziabicki, A., Fundamentals of Fibre Formation: The Science of Fibre Spinning and
Drawing, John Wiley & Sons, London, 1976.
32. Hancock, T.A., Spruiell, J.E., and White, J.L., Wet spinning of aliphatic and aromatic
polyamides, J. Appl. Polym. Sci., 21, 1227, 1977.
33. Hancock, T.A., An Experimental and Theoretical Study of the Wet Spinning Process,
Ph.D. Thesis, Department of Chemical Engineering, University of Tennessee, avail-
able through UMI, 1981.
34. Chenevey, E.C. and Wadhwa, L.H., Technical Report AFML-TR-82-4194, Air Force
Materials Laboratory, Dayton, OH, 1982.
35. Berry, G.C., Wong, C.P., Venkatamen, S., and Chu, S.G., Technical Report: AFML-
TR-79-4115, Air Force Materials Laboratory, Dayton, OH, 1979.
36. Pottick, L.A., The Inuence of Drying on the Structure and Mechanics of Poly
(P-Phenylene Benzobisthiazole) Fibers. Ph.D. Thesis, Department of Polymer
Science and Polymer Engineering, University of Massachusetts, available through
UMI, 1986.
37. Davis, V.A., Ericson, L.M., Parra-Vasquez, A.N.G., Fan, H., Wang, Y., Prieto, V., Longoria,
J.A., Ramesh, S., Saini, R., Kittrell, C., Billups, W.E., Adams, W.W., Hauge, R.H.,
Smalley, R.E., and Pasquali, M., Phase Behavior and Rheology of SWNTs in
Superacids, Macromolecules, 37, 154160, 2004.
38. Ramesh, S., Ericson, L.M., Davis, V.A., Saini, R.K., Kittrell, C., Pasquali, M., Billups,
W.E., Adams, W.W., Hauge R.H., and Smalley, R.E., Dissolution of pristine single
walled carbon nanotubes in superacids by direct protonation, J. Phys. Chem. B, 108,
87948798, 2004.
39. Davis, V.A., Ericson, L.M., Saini, R., Sivarajan, R., Hauge, R.H., Smalley, R.E., and
Pasquali, M., Proceedings of the 2001 AIChE Annual Meeting, Reno, NV, 351359,
2001.
40. Bahr, J.L., Mickelson, E.T., Bronikowski, M.J., Smalley, R.E., and Tour, J.M., Dis-
solution of small diameter single-wall carbon nanotubes in organic solvents?, Chem.
Commun., 2, 193194, 2001.
41. Georgakilas, V., Voulgaris, D., Vazquez, E., Prato, M., Guldi, D.M., Kukovecz, A.,
and Kuzmany, H., Purication of HiPco carbon nanotubes via organic functionaliza-
tion, J. Am. Chem. Soc., 124, 1431814319, 2002.
42. Saini, R.K., Chiang, I.W., Peng, H.Q., Smalley, R.E., Billups, W.E., Hauge, R.H.,
and Margrave, J.L., Covalent sidewall functionalization of single wall carbon nano-
tubes, J. Am. Chem. Soc., 125, 36173621, 2003.
43. Ying, Y.M., Saini, R.K., Liang, F., Sadana, A.K., and Billups, W.E., Functionalization
of carbon nanotubes by free radicals, Org. Lett., 5,14711473, 2003.
44. Poulin, P., Vigolo, B., and Launois, P., Films and bers of oriented single wall
nanotubes, Carbon, 40, 17411749, 2002.
45. Ericson, L.M., Macroscopic Neat Single-Wall Carbon Nanotube Fibers, Ph.D. Thesis
Department of Applied Physics, Rice University, available through UMI, 2003.
46. Kiss, G.D., Rheology and Rheo-Optics of Concentrated Solutions of Helical Polypep-
tides, Ph.D. Thesis, Department of Polymer Science and Polymer Engineering, Uni-
versity of Massachusetts, available through UMI, 1979.
47. Andrews, R.J., Rao, D., Rantell, A.M., Derbyshire, T., Chen, F., Chen, Y., J. Haddon,
R.C, Nanotube composite carbon bers, Appl. Phys. Lett., 75, 1329, 1999.
48. Haggenmueller, R., Gommans, H.H., Rinzler, A.G., Fischer, J.E., and Winey, K.I.,
Aligned single-wall carbon nanotubes in composites by melt processing methods,
Chem. Phys. Lett., 330, 219225, 2000.
49. Van Hattum, F.W.J., Bernardo, C.A., Finegan, J.C., Tibbetts, G.G., Alig, R.L., and Lake,
M.L., A study of the thermomechanical properties of carbon ber-polypropylene
composites, Polym. Composites, 20, 683688, 1999.
50. Tibbetts, G.G. and McHugh, J.J., Mechanical properties of vapor-grown carbon ber
composites with thermoplastic matrices, J. Mater. Res., 14, 28712880, 1999.
51. Lozano, K., Bonilla-Rios, J., and Barrera, E.V., A study on nanober-reinforced
thermoplastic composites (II): investigation of the mixing rheology and conduction
properties, J. Appl. Polym. Sci., 80, 11621172, 2001.
52. Ma, H.M., Zeng, J.J., Realff, M.L., Kumar, S., and Schiraldi, D.A., Processing,
structure, and properties of bers from polyester/carbon nanober composites, Com-
pos. Sci. Technol., 63, 16171628, 2003.
53. Bhattacharyya, A.R., Sreekumar, T.V., Liu, T., Kumar, S., Ericson, L.M., Hauge, R.H.,
and Smalley, R.E., Crystallization and orientation studies in polypropylene/single
wall carbon nanotube composite, Polymer, 44, 23732377, 2003.
54. Chiang, I.W., Brinson, B.E., Huang, A.Y., Willis, P.A., Bronikowski, M.J., Margrave,
J.L., Smalley, R.E., and Hauge, R.H., Purication and characterization of single-wall
carbon nanotubes (SWNTs) obtained from the gas-phase decomposition of CO
(HiPco process), J. Phys. Chem. B, 105, 82978301, 2001.
55. Kitagawa, T., Ishitobi, M., and Yabuki, K., An analysis of deformation process on
poly-p-phenylenebenzobisoxazole ber and a structural study of the new high-modulus
type PBO HM+ ber, J. Polym. Sci. B, 38, 16051611, 2000.
56. Kumar, S., Dang, T.D., Arnold, F.E., Bhattacharyya, A.R., Min, B.G., Zhang, X.F.,
Vaia, R.A., Park, C., Adams, W.W., Hauge, R.H., Smalley, R.E., Ramesh, S., and
Willis, P.A., Synthesis, structure, and properties of PBO/SWNT composites, Macro-
molecules, 35, 90399043, 2002.
57. Balberg, I., Anderson, C.H., Alexander, S., and Wagner, N., Excluded volume and
its relation to the onset of percolation, Phys. Rev. B, 30, 39333943, 1984.
58. Sreekumar, T.V., Liu, T., Min, B.G., Guo, H., Kumar, S., Hauge, R.H., and Smalley,
R.E., Polyacrylonitrile single-walled carbon nanotube composite bers, Adv. Mater.,
16, 5861, 2004.
59. Liu, T., Sreekumar, T.V., Kumar, S., Hauge, R.H., and Smalley, R.E., SWNT/PAN
composite lm-based supercapacitors, Carbon, 41, 24402442, 2003.
60. Poulin, P., Vigolo, B., Penicaud, A., and Coulon, C., U.S. Patent, 2002/0102585 A1,
Method for Obtaining Macroscopic Fibres and Strips from Colloidal Particles and in
Particular Carbon Nanotubes, 2003.
61. Vigolo, B., Penicaud, A., Coulon, C., Sauder, C., Pailler, R., Journet, C., Bernier, P.,
and Poulin, P., Macroscopic bers and ribbons of oriented carbon nanotubes, Science,
290, 13311334, 2000.
62. Vigolo, B., Poulin, P., Lucas, M., Launois, P., and Bernier, P., Improved structure and
properties of single-wall carbon nanotube spun bers, Appl. Phys. Lett., 81,
12101212, 2002.
63. Lobovsky, A., Matrunich, J., Kozlov, M., Morris, R.C., Baughman, R.H., and Zakhidov,
A.A., U.S. Patent, 2002/0113335 A1, Spinning, Processing, and Applications of
Carbon Nanotube Filaments, Ribbons and Yarns, 2002.
64. Dalton, A.B., Collins, S., Munoz, E., Razal, J.M., Ebron, V.H., Ferraris, J.P., Coleman,
J.N., Kim, B.G., and Baughman, R.H., Super-tough carbon-nanotube bresthese
extraordinary composite bres can be woven into electronic textiles, Nature, 423,
703703, 2003.
65. Israelchvili, J.N., Intermolecular and Surface Forces, 2nd ed., Academic Press, London,
1992, chaps. 11 and 14.
66. Hiemenz, P.C. and Rajagopalan, R., Principles of Colloid and Surface Chemistry 2nd
Edition, Marcel Dekker, New York, 1997, chap. 11.
67. Neimark, A.V., Ruetsch, S., Kornev, K.G., and Ravikovitch, P.I., Hierarchical pore
structure and wetting properties of single- wall carbon nanotube bers, Nano Lett.,
3, 419423, 2003.
68. Nativ-Roth, E., Levi-Kalisman, Y., Regev, O., and Yerushalmi-Rozen, R., On the route
to compatibilization of carbon nanotubes, J. Polym. Eng., 22, 353368, 2002.
69. Barisci, J.N., Tahhan, M., Wallace, G.G., Badaire, S., Vaugien, T., Maugey, M., and
Poulin, P., Properties of carbon nanotube bers spun from DNA-stabilized disper-
sions, Adv. Funct. Mater., 14, 133138, 2004.
70. Dalton, A.B., Collins, S., Razal, J., Munoz, E., Ebron, V.H., Kim, B.G., Coleman,
J.N., Ferraris, J.P., and Baughman, R.H., Continuous carbon nanotube composite
bers: properties, potential applications, and problems, J. Mater. Chem., 14, 13,
2004.
71. Smalley, R.E., Saini, R., Sivarajan, R., Hauge, R.H., Davis, V.A., Pasquali, M.,
Ericson, L.M., U.S. Patent, 2003/0170166A1, Fibers of Aligned Single-Wall Carbon
Nanotubes and Process for Making the Same, 2003.
72. Ericson, L., Fan, H., Peng, H., Davis, V., Zhou, W., Sulpizio, J., Wang, Y., Booker,
R., Vavro, J., Guthy, C., Parra-Vasquez, A., Kim, M., Ramesh, S., Saini, R., Kittrell,
C., Lavin, G., Schmidt, H., Adams, W., Billups, W., Pasquali, M., Hwang W., Hauge,
R., Fischer J., and Smalley, R., Macroscopic, neat, single-walled carbon nanotube
bers, Science, 305, 14471450, 2004.
73. Onsager, L., The effects of shape on the interaction of colloidal particles, Ann. N.Y.
Acad. Sci., 51, 627, 1949.
74. Flory, P.J., Phase equlibria in solutions of rod-like particles, Proc. Royal Soc. London,
Ser. A, 234, 1956.
75. Lee, C. S. and Yakobson, B.I., in preparation.
76. Song, W.H., Kinloch, I.A., and Windle, A.H., Nematic liquid crystallinity of multiwall
carbon nanotubes, Science, 302, 13631363, 2003.
77. Islam, M.F., Alsayed, A.M., Dogic, Z., Zhang, J., Lubensky, T.C., and Yodh, A.G.,
Nematic nanotube gels, Phys. Rev. Lett., 92, 2004.
78. Saini, R., Sivarajan, R., Hauge, R.H., Davis, V.A., Pasquali, M., Ericson, L.M., Kumar,
S., Veedu, S., U.S. Patent, 2003/0133865A1, Single-Wall Carbon Nanotube Alewives,
Process for Making and Compositions Thereof, 2003.
79. Wang, Y., Seed Crystals and Catalyzed Epitaxy of Single-Walled Carbon Nanotubes,
Ph.D. Thesis, Department of Applied Physics, Rice University, available through
UMI, 2004.
80. Walters, D.A., Casavant, M.J., Qin, X.C., Huffman, C.B., Boul, P.J., Ericson, L.M.,
Haroz, E.H., OConnell, M.J., Smith, K., Colbert, D.T., and Smalley, R.E., In-plane-
aligned membranes of carbon nanotubes, Chem.Phys. Lett., 338, 1420, 2001.
81. Vavro, J., M. C. Llaguno, J. E. Fischer, S. Ramesh, R. K. Saini, L. M. Ericson, V. A.
Davis, R. H. Hauge, M. Pasquali and R. E. Smalley, Thermoelectric power of p-doped
single wall carbon nanotubes and the role of phonon drag, Phys. Rev. Lett., 90, 065503,
2003.
12
Carbon Nanofiber
and Carbon Nanotube/
Polymer Composite
Fibers and Films
Han Gi Chae, Tetsuya Uchida, and
Satish Kumar
CONTENTS
12.1 Introduction
12.2 Vapor-Grown Carbon Nanofibers (VGCNFs) and Polymer
Composite Fibers
12.2.1 PBZT-VGCNF Composite Fibers
12.2.2 PP-VGCNF Composite Fibers
12.2.3 PET-VGCNF Composite Fibers
12.2.4 PMMA-VGCNF Composite Fibers
12.3 Carbon Nanotubes (CNTs) and Polymer Composite Fibers
12.3.1 PBO-SWNT Fibers
12.3.2 PAN-SWNT Composite Fibers and Films
12.3.2.1 PAN-SWNT Composite Fibers
12.3.2.2 Oxidative Stabilization of PAN-SWNT
Composite Fibers
12.3.2.3 PAN-SWNT Composite Films
12.3.3 Polyvinyl Alcohol (PVA)/SWNT Composite Fibers
and Films
12.3.3.1 PVA-SWNT Composite Fibers
12.3.3.2 PVA-SWNT Composite Films
12.3.4 PMMA-SWNT Composite Fibers
12.4 Additional Aspects of CNT-Polymer Composites
12.4.1 Cross-linking of SWNTs by Oxidation and Effect of Nitric
Acid Treatment
12.4.2 Crystallization Behavior of Polymer with SWNT
12.4.3 Effect of SWNT Exfoliation and Orientation
12.5 Polymer/SWNT Application (Supercapacitor)
Acknowledgments
References
12.1 INTRODUCTION
In the early nineteenth century, Herman Staudinger began to publish articles assert-
ing that polymeric molecules are practically endless chains held together by ordinary
chemical bonds.
1
Even though this idea of practically endless chains appeared
foreign to most chemists of that time, these concepts ultimately led to the develop-
ment of the synthetic polymer and fiber industry, resulting in revolutionary devel-
opments in the field, particularly during the period of 19301980.
Developments in nanoscale science and technology and their applications to
polymers and fibers suggest that we are at the initial stages of another revolutionary
period that will bring about polymeric and fibrous materials with properties not yet
seen. The field of polymer/carbon nanotube composites alone is expected to open
new vistas in materials properties. About 100 research papers were published on
polymer/carbon nanotube composites in the 19912000 period. However, the number
of publications in this field is growing at a rapid rate, with 80, 160, and over 300
publications in 2001, 2002, and 2003, respectively. The pace of this research activity
is only going to grow, as carbon nanotubes, and particularly single-wall carbon
nanotubes, become more affordable, and as methods are developed to synthesize or
separate SWNTs of specific chirality and diameter. Due to variation in electronic
structure, different types of nanotubes are likely to have different interactions with
a given polymer. There are many different solvents or solvent combinations (surfac-
tant/ water, organic solvents, inorganic acids, etc.) that can be used to disperse
nanotubes in polymers. And there are literally unlimited opportunities for nanotube
functionalization, and for chemical reaction with polymers. One can quickly see that
there are thousands of combinations (polymer/nanotube/solvents/chemical reactions)
on which to base experiments.
Despite the fact that nearly 500 research papers appeared on polymer carbon
nanotube composites in 20022003, the field is still in its infancy and is at the
same state where carbon fiber-based composites were in the 1960s when carbon
fibers were just being developed. Carbon fibers and their composite manufacturing
technology have matured after 40 years of development. In the past 40 years, carbon
fiber cost has also come down dramatically. From almost no composites in aero-
space structures in the 1960s, we have now reached the stage where 50% of the
weight of the Boeing 7E7 to be produced in 2008 will be composite materials.
There is sufficient scientific data to suggest that in 20 years, carbon nanotube-
enabled materials may reach similar application levels, with the added difference
that carbon nanotube-based materials will be much more functional. This chapter
summarizes recent developments in the field in the authors laboratory at Georgia
Institute of Technology.
12.2 VAPOR-GROWN CARBON NANOFIBERS
(VGCNFS) AND POLYMER COMPOSITE FIBERS
Carbon fibers (typical diameter 7 ~ 10 Pm) for composite applications were devel-
oped during the 1960s, while the development of carbon nanofibers (typical diameter
50 ~ 200 nm) dates back to the 1980s.
2
The primary interest of this section is the
VGCNF, which is synthesized from the pyrolysis of hydrocarbons or carbon mon-
oxide in the gaseous state in the presence of a catalyst.
24
VGCNFs distinguish
themselves from other types of carbon fibers, such as polyacrylonitrile- or mesophase
pitch-based carbon fibers, in their method of production, physical properties, and
structure. The reported moduli of VGCNFs are varied in a wide range from less
than 100 up to 700 GPa.
58
This noticeable variance in mechanical properties orig-
inates from the characteristic morphology of VGCNFs. High-resolution transmission
electron microscopy (HRTEM) of VGCNFs revealed two types of morphologies.
9
In one case, a truncated cone microstructure was seen, with outer and inner diameters
of 60 and 25 nm, respectively. In this type of VGCNF, graphite sheets were oriented
at an average angle of about 15q with respect to the fiber axis. Figure 12.1 and
FIGURE 12.1 Transmission electron micrographs of single-layer carbon nanofibers.
9
Fig. 12.2 show the TEM and HRTEM images of single-layer VGCNF. The HRTEM
image shows graphite layer stacking as well as graphite layer folding.
The second type of VGCNF is a double-layer carbon nanofiber, with outer and
inner diameters of 85 and 20 nm, respectively. Truncated cone structure is also
observed in the double-layer VGCNF as shown in Fig. 12.3.
The focus of VGCNF-reinforced composites has been the engineering applica-
tions that require superior strength, stiffness, and electrical and thermal conductivities.
FIGURE 12.2 High-resolution transmission electron micrograph of single-layer carbon
nanofiber.
9
FIGURE 12.3 Transmission electron micrograph of double-layer carbon nanofiber.
9
The matrices have included a variety of commodity polymers, including polypropy-
lene (PP),
1012
polycarbonate,
13,14
nylon,
15
polymethyl methacrylate (PMMA),
16
poly(ether ether ketone,
17
polyamide,
18
and epoxies.
19
The next sections highlight the
properties of a few VGCNFs containing polymers.
12.2.1 PBZT-VGCNF COMPOSITE FIBERS
20
A dope of poly(p-phenelyene benzobisthiazole) (PBZT)-VGCNF composite was
prepared by in situ polymerization in polyphosphoric acid (PPA) in the presence of
VGCNFs, and fibers were spun by dry-jet wet spinning. The degrees of dispersion
and orientation of the VGCNFs were assessed using transmission electron micros-
copy (TEM). The TEM image in Fig. 12.4 shows that the VGCNFs were isolated
without aggregation and were well oriented along the fiber axis. The lengths of the
VGCNFs appeared to be in the order of a few micrometers, that is, nearly the same
as pristine VGCNFs, which suggests that the PPA used in the in situ polymerization
of PBZT did not affect the geometry of the VGCNFs.
12.2.2 PP-VGCNF COMPOSITE FIBERS
21
Carbon nanofibers (5 wt%) and polypropylene (PP) composite fibers were spun
using conventional melt spinning.
FIGURE 12.4 Transmission electron micrograph of PBZT-CNF composite fiber.
20
Carbon
nano fiber
PBZT
200 nm
The tensile and compressive properties of the spun fibers are shown in Table 12.1.
The reinforcement effects of VGCNFs are evident from significant improvements
in tensile modulus as well as in compressive strength. Scanning electron microscopy
(SEM) and laser scanning confocal microscopy (LSCM) were employed to study
the microstructures of the composite fibers. The SEM image in Fig. 12.5 shows
good dispersion of VGCNFs in the matrix. In addition, the confocal microscope
images (not shown here) revealed an interesting phenomenon; that is, the outer layer
of the composite fiber contained fewer VGCNFs while having a higher degree of
orientation, as compared to the core.
12.2.3 PET-VGCNF COMPOSITE FIBERS
22
Polyethylene terephthalate (PET)-VGCNF composite fibers were also processed by
melt spinning. Figure 12.6 provides an SEM image of a PET-VGCNF composite
before fiber spinning, showing good VGCNF dispersion.
The mechanical properties of the pristine and the 5 wt% VGCNFs containing
composite fibers are given in Table 12.2. The tensile moduli of composite fibers are
slightly higher than those of the pristine PET fiber. On the other hand, the tensile
strengths of composite fibers are either comparable to or lower than the control fiber.
A series of process steps, including mixing, melt blending, and fiber spinning tended
to shorten the VGCNFs, which was detrimental to tensile properties. In addition,
TABLE 12.1
Mechanical Properties of PP and PP-CNF (95-5) Composite Fibers
21
Sample
Tensile Modulus
(GPa)
Tensile Strength
(GPa)
Strain to Failure
(%)
Compressive Strength
(MPa)
PP 4.6 r 0.7 0.49 r 0.06 23r 5 25 r 1
PP + CNF 7.1 r 0.9 0.57 r 0.07 16 r 2 48 r 10
FIGURE 12.5 Scanning electron micrograph of PP-CNF composite fiber.
21
the wide-angle x-ray diffraction (WAXD) study (not shown here) suggested that
significant misalignment of graphitic planes in the VGCNFs with respect to the
nanofiber axis direction may be another factor contributing to the relatively low
modulus of the PET-VGCNF composite fiber. The low orientation of the graphite
plane is consistent with the graphite plane misorientation (Fig. 12.2).
9
While only
moderate improvement in tensile modulus was observed, the composite fiber prop-
erties in compression and torsion improved by as much as 50% compared to the
control PET fiber (Table 12.2). Kozey et al.
23
reported that compressive failure in
polymeric fibers occurs by yielding, which results in the formation and propagation
of kinks. VGCNFs act as barriers for kink propagation, thus improving the com-
pressive strength.
FIGURE 12.6 Scanning electron micrograph of PET-CNT composite before fiber spinning.
22
TABLE 12.2
Mechanical Properties of PET and PET-CNF (95-5) Composite Fibers
22
Sample
Fiber
Diameter
(m)
Tensile
Modulus
(GPa)
Tensile
Strength
(GPa)
Strain
to Failure
(%)
Compressive
Strength
(MPa)
Torsional
Modulus
(GPa)
PET 24 r 3 10 r 3 0.43 r 0.08 26 r 2 80 0.7
PET/PR-24-PS 27 r 6 11 r 2 0.33 r 0.06 18 r 7 120 1.0
PET/PR-21-PS 27 r 5 11 r 3 0.25 r 0.1 13 r 8 100 0.9
PET/PR-24-HT 25 r 2 11 r 2 0.42 r 0.06 20 r 4 90 1.1
PR-24-PS, PR-21-PS, and PR-24-HT represent different fiber grades. PS: pyrolytically stripped;
HT: heat treated.
12.2.4 PMMA-VGCNF COMPOSITE FIBERS
24
Polymethyl methacrylate (PMMA) nanocomposites were processed by melt blend-
ing, containing two different grades of VGCNF at two different VGCNF loadings
(5 and 10 wt%). The PMMA-VGCNF composites were processed into 40-mm diam-
eter rods and 60-Pm diameter fibers. The respective microstructures and mechanical
and thermal properties were characterized.
The SEM image captured perpendicular to the extrusion direction and the frac-
tograph of the fracture surface (not shown here) of a PMMA-VGCNF composite
fiber showed good dispersion and bonding of the VGCNFs in the matrix. The mechan-
ical properties of pure PMMA and PMMA-VGCNF composite fibers are compared
in Table 12.3. Tensile modulus increased by more than 50%, and compressive
strength more than doubled by the addition of 5 wt% VGCNF. However, further
incorporation of VGCNFs up to 10 wt% did not result in any further improvements
in the mechanical properties. Tensile strength and elongation at break slightly
decreased with the addition of VGCNFs.
A series of thermal analysis experiments showed significant improvements
in the thermal behavior of the composite fiber. Thermogravimetric analysis
showed that the onset degradation temperature increased by about 50qC by
reinforcing PMMA with VGCNFs. In addition, dynamic mechanical analysis
revealed that the glass transition temperature of the PMMA-VGCNF composite
fiber is about 5qC higher than that of the control PMMA fiber. Finally thermo-
mechanical analysis showed only 5% of thermal shrinkage in the PMMA-CNF
(95-5) composite fiber at 110qC, while the control PMMA fiber shrank by 30%.
The improvement in thermal shrinkage was especially dramatic compared to other
thermal behavior because PMMA, a glassy polymer, tends to be subjected to
considerable shrinkage when the oriented fibers are allowed to relax at elevated
temperatures.
TABLE 12.3
Mechanical Properties of PMMA and PMMA-CNF Composite Fibers
24
Sample
Fiber
Diameter
(m)
Tensile
Modulus
(GPa)
Tensile
Strength
(GPa)
Strain
to Failure
(%)
Compressive
Strength
(MPa)
PMMA 60 r 4 4.7 r 1.5 0.20 r 0.04 16 r 3 28 r 2
PMMA/PR-21-PS
(955)
61 r 12 8.0 r 1.2 0.17 r 0.04 10 r 6 73 r 11
PMMA/PR-21-PS
(9010)
63 r 10 7.7 r 1.0 0.16 r 0.04 10 r 6 N/A
PMMA/PR-24-PS
(955)
62 r 5 7.5 r 1.3 0.16 r 0.03 10 r 5 66 r 20
PMMA/PR-24-PS
(9010)
63 r 5 7.6 r 0.9 0.15 r 0.01 9 r 4 N/A
12.3 CARBON NANOTUBES (CNTS) AND POLYMER
COMPOSITE FIBERS
Carbon nanotubes (CNTs), which were discovered in 1991,
25
are extremely versatile
materials; they are very strong, yet highly elastic, highly conducting, and nanoscale
in size, but stable and robust in most chemically harsh conditions.
26
As a result,
many potential applications have been proposed for CNTs, including conductive and
high strength composites, energy storage devices, sensors, field emission displays,
nanoscale semiconductor devices, probes, and interconnects .
2630
Significant
research activity has focused on the development of a new class of high-performance
polymer composites with nanotubes as the reinforcing fillers.
While carbon fibers, carbon nanofibers, multiwall carbon nanotubes (MWNTs),
and single-wall carbon nanotubes (SWNTs) are being used for reinforcing polymer
matrices, SWNTs are sometimes heralded as the ultimate in carbon reinforcement.
Therefore, this section will mostly discuss SWNT-polymer composite fibers and
their potential.
12.3.1 PBO-SWNT FIBERS
31
Poly(p-phenylene benzobisoxazole) (PBO) was in situ polymerized in the presence
of single-wall carbon nanotubes (SWNTs) in polyphosphoric acid (PPA) using
typical PBO polymerization conditions. PBO and PBO-SWNT lyotropic liquid crys-
talline solutions in PPA were spun into fibers using dry-jet wet spinning. The tensile
strength of the PBO-SWNT fiber containing 10 wt% SWNT was over 50% higher
than those of the control PBO fibers containing no SWNT (Table 12.4).
Synthesis and fiber spinning experiments were performed at three levels of
impurity (30, 10, and 1 wt% metal catalytic impurity) in the SWNTs. At 30%
catalytic impurity, continuous fiber could not be drawn. At 10 wt% impurity, fiber
could be drawn; however, its mechanical properties were comparable to those of the
PBO synthesized and spun under the same conditions. Dramatic improvements in
tensile strength were obtained only when SWNTs with about 1 wt% impurity were
used. Extensive high-resolution transmission electron microscopy in PBO-SWNT
(9010) has not revealed SWNT aggregates or bundles, suggesting that SWNTs
TABLE 12.4
Mechanical Properties of PBO and PBO-SWNT Composite Fibers
31
Sample
Fiber
Diameter
(Pm)
Tensile
Modulus
(GPa)
Tensile
Strength
(GPa)
Strain
to Failure
(%)
Compressive
Strength
(GPa)
PBO 22 r 2 138 r 20 2.6 r 0.3 2.0 r 0.2 0.35 r 0.6
PBO-SWNT
(955)
25 r 2 156 r 20 3.2 r 0.3 2.3 r 0.3 0.40 r 0.6
PBO-SWNT
(9010)
25 r 2 167 r 15 4.2 r 0.5 2.8 r 0.3 0.50 r 0.6
were indeed well dispersed in these samples. On the other hand, both oriented and
unoriented nanotube bundles have been observed in PBO-SWNT samples containing
15 wt% nanotubes. Attempts are being made to achieve better SWNT dispersion at
this and higher concentrations.
12.3.2 PAN-SWNT COMPOSITE FIBERS AND FILMS
12.3.2.1 PAN-SWNT Composite Fibers
32
The tensile and storage moduli of polyacrylonitrile (PAN)-SWNT fibers containing
10 wt% SWNT increased by factors of two and ten at room temperature and at
150qC, respectively, when compared to the PAN fiber (Fig. 12.7(a)). In addition, the
glass transition temperature increased by 40qC (Fig. 12.7(b)).
Mechanical properties of a two-phase composite, in which both the matrix and
the reinforcement have different orientations,
33
require a full set of elastic constants
for the anisotropic matrix that are currently not available for PAN. Therefore the
composite fiber modulus (E
c
) was estimated using the equation,
E
c
= V
NT
E
NT
+ V
PAN
E
PAN
FIGURE 12.7 (a) storage modulus and (b) tan G as a function of temperature for PAN and
PAN-SWNT composite fibers.
32
where E
NT
and E
PAN
are the nanotube and PAN moduli along the ber axis, and V
NT
and V
PAN
are their volume fractions, respectively. Based on the SWNT orientation
factors for 5 and 10 wt% composite bers and the orientation dependence of the
SWNT modulus, E
NT
values obtained from elsewhere
34,35
for two different rope diam-
eters are listed in Table 12.5, along with values of E
PAN
, V
NT
, and V
PAN
. Table 12.5
shows that the experimental moduli values of PAN-SWNT composite ber are well
within the range predicted for the two rope diameters. Rope diameter for the SWNT
powder used in this study was measured to be 37 8 nm from scanning electron
microscopy.
36
However, the measured modulus is closer to the predictions based on
smaller diameter ropes. This suggests that some SWNT rope exfoliation has occurred
during mixing, ber processing, and ber drawing. Based on high resolution trans-
mission electron microscopy study (Fig. 12.8(a) and (b)), average single-wall nano-
tube bundle diameter in PAN-SWNT (95-5) ber was measured to be 11 nm,
conrming partial exfoliation.
37
Further improvement in nanotube exfoliation and
orientation is expected to result in further modulus increase.
32,35
Composite ber also showed signicant reduction in thermal shrinkage, as well as
in polymer solubility. For example, the shrinkage at nominal stress at 200C was approx-
imately 25% in PAN ber and only about 10% in PAN-SWNT (9010) composite ber,
as shown in Fig. 12.9. After extensive sonication and being subjected to elevated tem-
perature, less than 50% of the PAN could be dissolved from the PAN-SWNT (9010)
composite ber, thus making the composite ber substantially more solvent resistant.
These observations suggest good interaction between PAN and SWNTs.
SWNTs exhibited higher orientation than PAN as determined by Raman spec-
troscopy and x-ray diffraction. Infrared spectroscopy conrmed the anisotropic optical
absorption behavior of SWNTs, as shown in Fig. 12.10.
TABLE 12.5
Physical and Mechanical Properties of SWNT-PAN Composite
Fibers
32
PAN
PAN-SWNT
(955 wt%)
PAN-SWNT
(9010 wt%)
V
NT
0 4.6 9.2
f
PAN
(x-ray) 0.58 0.68 0.66
f
NT
(Raman) 0.94 0.92
E
PAN
(GPa)
7.9 9.4 9.1

E
NT
(GPa) (SWNT rope diameter > 20 nm)
*
28.2 22.2
E
NT
(GPa) (SWNT rope diameter ~ 4.5 nm)
*
149.1 122.6
E
c
(GPa) (SWNT rope diameter 20 nm) 10.3 10.3
E
c
(GPa) (SWNT rope diameter ~ 4.5 nm) 15.8 19.5
E
exp
(GPa) 7.9 14.2 16.2
V
NT
calculated assuming PAN and SWNT densities of 1.18 g/cm
3
and 1.30 g/cm
3
, respec-
tively. E
PAN
for the composite bers was calculated from the control PAN ber modulus
to account for the difference in orientation.
*
Obtained from Reference 35 for the given f
NT
and rope diameter.
FIGURE 12.8 Transmission electron micrographs of SWNT ropes in PAN: (a) tip of nano-
tube; (b) intertube distances.
37
FIGURE 12.9 Thermal shrinkage in (A) PAN, (B) PAN-SWNT (955), and (C) PAN-SWNT
(9010) composite fibers as a function of temperature.
32
5
0
5
10
15
20
25
C
h
a
n
g
e

i
n

l
e
n
g
t
h

(
%
)
0 50 100 150 200 250
Temperature (C)
PAN fiber
PAN/SWNT (5/5) fiber
PAN/SWNT (90/10) fiber
12.3.2.2 Oxidative Stabilization on PAN-SWNT
Composite Fibers
38
The work on PAN-SWNT composites is also important because PAN is a carbon
fiber precursor and PAN-SWNT composite fiber is likely to result in higher tensile
strength and modulus carbon fibers for aerospace applications. Polyacrylontrile,
39,40
petroleum pitch, and cellulosic fibers have been used as carbon fiber precursors.
Currently carbon fibers are predominantly made from polyacrylonitrile copolymer
precursors.
41
For converting PAN to carbon fiber, thermo-oxidative stabilization
typically in the 200 to 300qC range is a key step.
At 10 wt% SWNTs, breaking strength, modulus, and strain to failure of the
oxidized composite fiber increased by 100, 160, and 115%, respectively (Table 12.6).
Tensile fracture surfaces (not shown here) of thermally stabilized PAN and the PAN-
SWNT fibers exhibited brittle behavior, and well-distributed SWNT ropes covered
with the oxidized matrix can be observed in the tensile fracture surfaces of the fibers.
Figure 12.11 shows weight loss in PAN and PAN-SWNT composite fibers when
isothermally oxidized at 250qC. Weight loss during oxidative stabilization of PAN
is a well-known phenomenon, even though there is oxygen uptake.
42
One reason for
the difference in weight loss between the control PAN and the PAN-SWNT com-
posite fibers is that SWNTs are thermally stable at this temperature. However, the
mass balance analysis suggests that in the oxidized PAN-SWNT (9010) composite
fiber, the weight residue is about 0.3 wt% more than the calculated value, assuming
that weight loss in PAN in the composite fiber is the same as in the control PAN
and that there was no weight loss in SWNT. This suggests that the presence of
SWNT may alter PAN oxidative stabilization rate, mechanism, or both.
FIGURE 12.10 Polarized IR spectra of PAN and PAN-SWNT (991) composite fibers.
32
100
80
60
40
20
0
T
r
a
n
s
m
i
s
s
i
o
n

(
%
)
2000 2100 2200 2300 2400 2500
Wave number (cm
1
)
PAN/SWNT (II)
PAN (II and )
PAN/SWNT ( )
PAN 0-degree
PAN 90-degree
PAN/1% SWNT 0-degree
PAN/1% SWNT 90-degree
The tensile strength and modulus of carbon fibers strongly depend on the struc-
ture and properties of the PAN precursor fiber. As listed in Table 12.6, the mechanical
properties of the pristine and stabilized composite fibers were higher than those of
the control PAN fiber. In addition, PAN had higher orientation in the composite fiber
than in the control PAN fiber, and composite fiber exhibited reduced stress during
stabilization. In the 100 to 1060 GPa modulus range, carbon fiber modulus depended
very strongly on the orientation. With a misorientation factor of less than 5q, graphite
in plane modulus dropped from 1060 GPa to below 200 GPa.
43
Carbon fibers used
in the aerospace industry today have a tensile modulus in the 200 to 300 GPa range
TABLE 12.6
Mechanical Properties of PAN and PAN-SWNT Composite Fibers before
and after Oxidation
38
SWNT Content
(wt%)
Tensile Modulus
(GPa)
Tensile Strength
(GPa)
Strain to Failure
(%)
Before oxidation 0
5
10
7.9 r 0.4
14.2 r 0.6
16.2 r 0.8
0.23 r 0.03
0.36 r 0.02
0.33 r 0.02
11.6 r 1.4
11.9 r 1.3
9.7 r 1.6
After oxidation 0
5
10
5.2 r 0.5
8.9 r 1.0
13.5 r 0.9
0.06 r 0.01
0.10 r 0.01
0.13 r 0.01
1.4 r 0.3
2.2 r 0.5
3.1 r 0.7
FIGURE 12.11 Weight loss in PAN and PAN-SWNT composite fibers when isothermally
heated at 250qC under air flow.
38
100
98
96
94
92
90
0 20 40 60 80 100 120 140 160
Time (min)
300
250
200
150
100
50
0
T
e
m
p
e
r
a
t
u
r
e

(
C
)
W
e
i
g
h
t

(
%
)
Neat PAN
PAN/SWNT (95/5)
PAN/SWNT (90/10)
Temperature
and are carbonized at about 1500qC. Although PAN-based carbon fibers with moduli
higher than 500 GPa can be produced, these fibers are carbonized at ~2500qC and have
low compressive strength (~1 GPa). By comparison, PAN-based carbon fibers carbon-
ized at ~1500qC with a modulus in the 200 to 300 GPa range have high compressive
strength values (~3 GPa) .
44
All pitch-based carbon fibers have low compressive strength
(less than 1 GPa).
44,45
Significant improvement in carbon fiber modulus without
sacrificing the compressive strength will represent a major development in the carbon
fiber composite materials technology. Fiber compressive strength decreases with
increasing carbonization temperature and with increasing graphitic order. The key
to increasing the carbon fiber modulus without sacrificing compressive strength is
to achieve higher orientation at moderate carbonization temperatures (~1500qC).
Since PAN-SWNT composites exhibit lower stress during stabilization and have
higher PAN orientation, PAN-SWNT fibers may yield higher modulus carbon fibers
while maintaining compressive strength. In PAN-SWNT fibers reported here, SWNT
ropes were partially exfoliated. The rope diameter was reduced from 30 nm to 11
nm. Exfoliation to individual nanotube will result in fiber property improvements
at lower SWNT concentration than have been used in the current work, thus reducing
the nanotube requirements, which should make these composite systems a commer-
cial reality.
12.3.2.3 PAN-SWNT Composite Films
46
The mechanical properties of PAN-SWNT composite films listed in Table 12.7
show that the initial modulus of the composite film is up to 2.7 times the modulus
of PAN film, and more than eight times the modulus of the bucky paper. Tensile
strength of the composite film is up to twice the strength of the PAN film, and ten
times the strength of the bucky paper. This clearly demonstrates the reinforcing
effect of PAN in SWNTs and vice versa. The tensile strength and modulus values
of the composite film with the unpurified SWNT are larger than those using the
TABLE 12.7
Mechanical and Electrical Properties of PAN, PAN-SWNT, and SWNT
Films
46
Sample
SWNT
Content
(wt%)
Tensile
Modulus
(GPa)
Tensile Yield
Strength
(MPa)
Strain
to Failure
(%)
Conductivity
(S/m)
PAN 0 2.7 r 0.4 57 r 4 22.3 r 6.5 N/A
PAN/SWNT (80/20)
a
20 6.4 r 0.7 78 r 10 1.9 r 0.4 2.4 u10
3
PAN/SWNT (60/40)
a
40 7.2 r 1.0 65 r 22 2.0 r 0.6 6.3 u10
3
PAN/SWNT (60/40)
b
40 10.9 r 0.3 103 r 18 1.6 r 0.6 1.5 u10
4
SWNT bucky paper 100 0.8 r 0.1 10 r 2 5.6 r 0.3 3.0 u 10
4
a
Purified HiPCO SWNTs were used to prepare composite film.
b
Unpurified HiPCO SWNTs were used to prepare composite film.
purified SWNTs at the same SWNT loading. Due to the presence of 24 nm diameter
catalytic particles, unpurified SWNTs are easier to disperse in dimethylformamide
(DMF). Better mechanical properties of the composite film processed using unpuri-
fied SWNTs are attributed to the difference in dispersion quality between unpurified
and purified films. The mechanical properties of PAN-SWNT composites are signi-
ficantly higher than the SWNT bucky paper and control PAN film. In addition, the
electrical conductivities of PAN-SWNT films exhibited fairly comparable values,
which implies that PAN-SWNT composite films maintain the degree of percolation
as well as reinforce each other.
12.3.3 POLYVINYL ALCOHOL (PVA)/SWNT
COMPOSITE FIBERS AND FILMS
12.3.3.1 PVA-SWNT Composite Fibers
47
The concept of producing high-performance fibers from flexible polymers is based
on arresting polymer orientation and preventing chain folding by gel spinning, which
induces crystallization and minimizes chain entanglements. High mechanical
drawability is the key for obtaining high orientation. In order to obtain high drawabil-
ity, entanglement control between the chains is crucial. A certain degree of entan-
glement is needed to endure drawing, but too much entanglement would reduce
drawability. In other words, the concentration of the polymer solution should be
near the polymer overlap concentration. But, in reality, much higher concentration
is chosen as fibers cannot be drawn near the overlap concentration. For gel spinning
of PVA or other polymers, crystallization-induced gelation is the key step in the
process. These crystals act as physical cross-links for mechanical drawing. For good
drawability, the crystal size should be as small as possible.
The PVA-SWNT dispersions in DMSO and DMSO-H
2
O solvents were prepared
and successfully spun into composite fibers through gel spinning. Mechanical prop-
erties of gel-spun PVA and PVA-SWNT composite fibers are listed in Table 12.8.
12.3.3.2 PVA-SWNT Composite Films
48
The ultimate reinforcement of fillers can be achieved only when the applied stress
to matrix polymer is transferred into dispersed fillers. Intensive study for Raman
spectroscopy of CNTs indicated that D
*
(disorder) band of CNTs can represent the
deformation of CNTs from external stresses. Figure 12.12 and Fig. 12.13 show the
TABLE 12.8
Mechanical Properties of PVA and PVA-SWNT Gel-Spun Fibers
47
Sample
Fiber Diameter
(Pm)
Tensile Strength
(GPa)
Tensile Modulus
(GPa)
Strain to Failure
(%)
PVA 26.5 r 1.6 0.9 r 0.1 25.6 r 2.6 7.5 r 1.6
PVA/SWNT (3 wt%) 27.0 r 2.0 1.1 r 0.2 35.8 r 3.5 8.8 r 1.7
stress-strain curves of PVA-SWNT films and D
*
band shift versus strain, respectively.
The shifting behavior of the D
*
band looks similar to that of stress variation with
strain, which implies the load transfer to CNTs from polymer matrix.
12.3.4 PMMA-SWNT COMPOSITE FIBERS
49
Liu et al.
50
showed the wrapping behavior of polymer around SWNT. Figure 12.14
shows the SEM image of electrospun PMMA-SWNT composite fiber. SWNT is elec-
trically conductive, while PMMA is an insulating matrix around SWNT. Therefore this
FIGURE 12.12 Stress-strain curves for various films.
48
FIGURE 12.13 Raman D
*
band peak position as a function of strain for PVA-PVP-SDS-SWNT
composite films containing 5 wt% SWNT. Excitation laser wavelength was 785 nm, and the
laser power was 1.5 mW.
48
150
100
50
0
0 10 20 30 40 50 60 80 190 200 210
Strain (%)
T
e
n
s
i
l
e

s
t
r
e
e
s

(
M
P
a
)
PVA/PVP/SDS/SWNT
PVA/PVP/SDS
PVA
2566
2568
2570
2572
2574
2576
2578
2580
2582
0 10 20 30 40 50
Strain (%)
D

b
a
n
d

p
e
a
k

p
o
s
i
t
i
o
n

(
c
m
1
)
type of fiber can be used as nanowire in nanoscale electrical devices. Such fibers
can also be used as atomic force microscope (AFM) tips.
51
12.4 ADDITIONAL ASPECTS
OF CNT-POLYMER COMPOSITES
12.4.1 CROSS-LINKING OF SWNTS BY OXIDATION AND EFFECT
OF NITRIC ACID TREATMENT
52,53
It has been proposed that mechanical properties of SWNT products can be improved
by covalent cross-linking.
5456
Oxidation of CNTs is one possible way to achieve
cross-linking. Functional groups such as carboxylic acid, quinine,
57
phenol,
58
ester,
59
amide,
60
and zwitterions
61
have been reported on the oxidized SWNTs. Oxidation
can also occur during SWNT purification in strong acid.
62,59
It is understood that
oxidation occurs at end caps and at the defect sites.
63
The presence of polycyclic
aromatic amorphous carbon has been reported in the oxidized SWNT.
53,64
Small-
diameter SWNTs have been reported to be destroyed during nitric acid treat-
ment,
53,65
and the formation of continuous phase morphology has been observed in
HNO
3
-purified SWNTs, which is attributed to amorphous carbon formed by the
decomposition of SWNTs.
66
As listed in Table 12.9, the nitric acid-treated bucky paper showed higher
mechanical properties than oleum-treated bucky paper, which may be evidence of
cross-linking by oxidation of SWNTs. However, the in-plane electrical conductivity
decreases with nitric acid treatment, which is a result of SWNT functionalization
leading to imperfections and defects along the SWNT wall.
The storage modulus of SWNT films is fairly constant (Fig. 12.15) in the
measured temperature range (there is slight up-turn above 150qC) and the tan G value
FIGURE 12.14 Scanning electron micrograph of electrospun PMMA-SWNT fiber.
49
is very low (~0.02), suggesting that these films are fairly elastic in this temperature
range.
It also has been reported that the most important parameter, that determines the
chemical oxidation attack of the nanotubes is their diameter. Small diameter tubes,
due to the stress induced by the curvature, are first attacked and destroyed. Some
reactivity is also reported for the larger diameter metallic nanotubes.
67
From the
quantitative analysis of CNT diameter with nitric acid treatment, one can see that
the relative fractions of small diameter SWNTs (0.88 and 0.89 nm) significantly
decreased with increasing nitric acid concentration, as shown in Fig. 12.16. The
percentage of these small diameter tubes decreased from 70% in the control film
to less than 20% in 10 M samples, confirming the selective degradation of the small
TABLE 12.9
Mechanical and Electrical Properties of Various SWNT Films
53
Sample
Tensile Modulus
(GPa)
Tensile Strength
(MPa)
Strain to Failure
(%)
Conductivity
(S/m)
Control SWNT film
*
0.8 r 0.1 10 r 2 5.6 r 0.3 3.0 u 10
4
3M HNO
3
SWNT film 1.4 r 0.1 16 r 1 1.4 r 0.2 2.3 u 10
4
6M HNO
3
SWNT film 2.9 r 0.2 71 r 5 3.4 r 0.2 2.4 u 10
4
10M HNO
3
SWNT film 5.0 r 0.2 74 r 2 3.0 r 0.1 1.2 u 10
4
*
Control SWNT film is bucky paper processed from aqueous solution without treating nitric acid.
FIGURE 12.15 Dynamic mechanical behavior of SWNT film processed from 6 molar nitric
acid.
53
0.2 5
4
3
2
1
0
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
20 40 60
T
a
n
Tan
80 100 120 140 160 180 200 220
Temperature (C)
E

(
G
P
a
)
E
diameter SWNTs by HNO
3
. As a result, the relative fraction of the large diameter
SWNTs (1.19 and 1.11 nm) increased monotonically with nitric acid concentration.
It has been pointed out that the data in Fig. 12.16 are based on the Raman radial
breathing mode using 785-nm wavelength laser. Therefore metallic tubes have not
been observed in this study.
12.4.2 CRYSTALLIZATION BEHAVIOR OF POLYMER
WITH SWNT
67
The spherulite size in PP is much larger than in PP-SWNT, suggesting that SWNTs
act as nucleating sites for PP crystallization (Fig. 12.17(a) and (b)). Based on the
half-crystallization time as a function of isothermal crystallization temperature, it
has been shown that the addition of 0.8 wt% SWNT increases the crystallization
rate by as much as an order of magnitude or higher (Fig. 12.18). In the drawn PP-
SWNT composite fibers, SWNT orientation was monitored using Raman spectros-
copy and that of PP using x-ray diffraction. These studies show that the orientation
of SWNTs (orientation factor 0.95 assuming Gaussian distribution) was higher than
that of PP (orientation factor 0.86).
FIGURE 12.16 Relative fractional changes of different diameter SWNTs as a function of
HNO
3
concentration.
53
0.2
0.15
0.1
0.05
0
0 5 10
0 5 10
0.2
0.15
0.1
0.05
0
0 5 10
0.2
0.15
0.1
0.05
0
0 5 10
0.8
0.6
0.4
0.2
0
0 5 10
0.8
0.6
0.4
0.2
0
0 5 10
0.8
0.6
0.4
0.2
0
d = 1.19 nm d = 1.11 nm d = 1.07 nm
d = 1.02 nm d = 0.89 nm d = 0.88 nm
Concentration of HNO
3
(M)
R
e
l
a
t
i
v
e

f
r
a
c
t
i
o
n

o
f

S
W
N
T
s
12.4.3 EFFECT OF SWNT EXFOLIATION AND ORIENTATION
35,77
Modulus of single-wall carbon nanotube (SWNT) films and fibers, calculated using
continuum mechanics and the SWNT rope elastic constants, is consistent with the
experimentally measured moduli of these products. Shear modulus of 20-nm diameter
ropes is about 1 GPa, and that of 4-nm diameter ropes is about 6 GPa. The axial
tensile modulus of large diameter ropes drops precipitously with orientation, while
the smaller diameter rope modulus exhibits lower orientation dependence
(Fig. 12.19).
35
Therefore, in addition to orientation, for the ropes composed of
SWNTs with varying diameters and helicity, exfoliation also appears to be critical
in achieving high modulus SWNT films and fibers.
Due to the strong resonance-enhanced Raman scattering,
6971
polarized Raman
spectroscopy is commonly used for SWNT orientation determination.
7276
By fitting
the polarized Raman scattering intensity as a function of angles that the sample
makes with respect to the polarization direction of the incident exciting laser source,
FIGURE 12.17 Optical micrographs (with cross-polars) of (a) polypropylene (PP) and (b) PP-
SWNT composite with 0.8 wt% SWNT content.
67
(b)
100 m
( )
the full width at half-maximum (FWHM) of Gaussian or Lorentzian distribution func-
tion is generally taken as the measure of orientational order. While this approach is
mathematically simple, the assumed distribution function may not truly reflect the
orientation distribution of SWNTs. Therefore the orientational order of SWNTs would
FIGURE 12.18 Crystallization half-time (t
1/2
) of (a) PP and (b) PP-SWNT (99.20.8) as a
function of crystallization temperature.
67
FIGURE 12.19 Effect of SWNT rope orientation on the axial tensile modulus of SWNT
fiber.
35
1000
800
600
400
200
0
120 122 124 126 128 130 132 134
Crystallization temperature (C)
a
b
t
1
/
2
(
s
e
c
)
0 10 20 30 40 50 60 70 80 90
10
3
10
2
10
1
10
0
FWHM of Gaussian distribution function, F() (degrees)
G
12
= 19.5 GPa
G
12
= 6 GPa
G
12
= 1 GPa
A
v
e
r
a
g
e
d

m
o
d
u
l
u
s

<
E
x
>

(
G
P
a
)
not be quantified correctly. To rectify this problem, a generalized spherical harmonics
expanded orientation distribution function (ODF) has been used to determine the SWNT
orientation.
77
A relationship between the second and the fourth order SWNT
orientation parameters and the Raman scattering intensity and depolarization ratio
has been established. By determining the VV and VH (direction of polarizer and
analyzer; V and H indicate vertical and horizontal, respectively) intensities at 0 and
90q to the orientation direction, the second and the fourth order SWNT orientation
parameters can be determined. From the second and the fourth order orientation param-
eters, orientation distribution function can be constructed using the maximum entropy
formalism.
12.5 POLYMER/SWNT APPLICATION
(SUPERCAPACITOR)
36
The diameter of the as-produced HiPco SWNT ropes used in this study was 35 r
7 nm and that for the as-prepared SWNT-PAN composite film was estimated to be
57 r 8 nm (Fig. 12.20(a) and (b)). The increased rope diameter indicates PAN
copolymer adsorption by the SWNT ropes. Physical or chemical activation of PAN
or its copolymers is generally used for producing activated carbon with high specific
surface area and porosity, and it is expected that similar activation treatment of
SWNT-PAN composite film can also be used for developing SWNT activated carbon
composite films.
The specific capacitance of the SWNT activated carbon film is significantly higher
than that of the SWNT bucky paper (Fig. 12.21). The specific capacitance of SWNT
activated carbon at 0.001 ampere discharging current is strongly dependent on the
discharging voltage, a result of its nonlinear discharging behavior. When the dis-
FIGURE 12.20 Scanning electron micrographs of (a) as-produced HiPco SWNT powder and
(b) as-produced SWNT-PAN composite film.
36
charging voltage is reduced from 0.7 volts to 0.2 volts, the specific capacitance
increased from 60 F/g to 380 F/g. The Ragone plot (Fig. 12.22) shows that both the
power and energy densities of SWNT-activated carbon composite film are signifi-
cantly higher than that of SWNT bucky paper.
FIGURE 12.21 Specific capacitance as a function of discharging voltage.
36
FIGURE 12.22 Ragone plot for various electrodes evaluated at discharging current of 0.01A.
36
400
350
300
250
200
150
100
50
0
0.7 0.6 0.5 0.4 0.3 0.2
Discharging voltage (V)
S
p
e
c
i
f
i
c

c
a
p
a
c
i
t
a
n
c
e

(
F
/
g
)
SWNT/activated carbon at 0.001A
SWNT/activated carbon at 0.01A
SWNT bucky paper at 0.001A
SWNT bucky paper at 0.01A
10
1
10
1
10
0
10
0
10
1
10
1
Energy density (kJ/kg)
SWNT/activated carbon film
Heat treated SWNT/PAN film
SWNT bucky paper
P
o
w
e
r

d
e
n
s
i
t
y

(
k
W
/
k
g
)
This chapter summarizes studies of a number of polymeric systems containing carbon
nanofibers as well as carbon nanotubes. Limited studies have also been reported on
carbon nanotube materials containing no polymers. Composites have been processed
using melt processing, solution processing, and in situ polymerization. Carbon
nanotubes containing polymeric materials exhibit improved tensile strength, tensile
modulus, strain to failure, torsional modulus, compressive strength, glass transition
temperature, solvent resistance, and reduced shrinkage. In polyacrylonitrile containing
10 wt% SWNT, the modulus above glass transition temperature is more than an order
of magnitude higher than the modulus of the control PAN fiber at that temperature.
Theoretical studies suggest that with SWNT exfoliation and with improved orientation,
similar property improvements can be expected at much lower SWNT content, making
these materials affordable. It is noted that all the property improvements are not
observed in every system. In addition, carbon nanotube-based materials will also have
electrical and thermal properties. These materials will also find applications for hydro-
gen storage and various types of sensors such as gas, strain or stress, and heat, among
others. Developments in production technology for carbon nanofibers, multiwall nano-
tubes (MWNT), double-wall nanotubes (DWNT), and single-wall nanotubes continue.
Between 0.4 and 4 nm, there can be nearly 400 different types of SWNTs. A given
production run today may contain about 20 different types of nanotubes, typically in
the 0.7 to 1.4-nm diameter range. Synthesis of a specific diameter and chirality SWNT
will represent a major breakthrough in the field and will unfold materials development
that will be considered revolutionary, even by the standards of the material develop-
ments that have occurred in the twentieth century. This situation perhaps can be
compared to the development of stereo-specific polymers that instantly made simple
polymers such as polypropylene much more useful. Professor Smalley
76
has pointed
out that there are only three types of polypropylene (atactic, syndiotactic, and isotactic),
but there are nearly 400 possibilities for the SWNTs. This points to both the enormous
potential of nanotubes and the complexity of the problem.
As we think about the future of potential materials development in the field, it
will help to keep in mind that even visionaries and highly successful scientists and
business people sometimes fail to comprehend and predict what is just around the
corner, as can be seen from the following quotations:
Radio has no future; x-rays are a hoax.Lord Kelvin (~1885).
I have not the smallest molecule of faith in aerial navigation other than
ballooning.Lord Rayleigh (1889).
I think there is a world market for about five computers.Thomas J. Watson
(IBM founder, 1943).
Having said this, the question to ponder is, what type of fiber development can
we expect in the coming century? Will we develop fibrous materials with the
following characteristics?
Ability to harvest and store the energy of the sun
Functionality of a computer and cell phone
Heating and cooling capability as needed
Production of light on demand
Ability to monitor health
Color and feel of a typical textile
ACKNOWLEDGMENTS
Funding from the Office of Naval Research (N00014-01-1-0657), Air Force Office
of Scientific Research (F49620-03-1-0124), National Science Foundation, and Car-
bon Nanotechnologies, Inc. is gratefully acknowledged.
REFERENCES
1. Hounshell, D.A. and Smith, J.K., Jr., Science and Corporate Strategy: Du Pont R&D,
19021980, Cambridge University Press, 1988, p. 231.
2. U.S. Patent 4,565,684 (January, 21, 1986), G. G. Tibbetts and M. G. Devour (to
General Motors Corporation).
3. Lake, M.L. and Ting, J.-M., in Burchell, T.D., Ed., Carbon Materials for Advanced
Technologies, Pergamon, Oxford, U.K., 1999, pp. 139167.
4. de Jong, D.P. and Geus, J.W., Carbon Nanofibers: Catalytic Synthesis and Applica-
tions, Catal. Rev. Sci. Eng., 42, 4, 481, 2000.
5. Jacobson, R.L., Tritt, T.M., Guth, J.R., Ehrlich, A.C., and Gillespie, D.J., Mechanical
Properties of Vapor-grown Carbon Fiber, Carbon, 33, 9, 1217, 1995.
6. Ting, J., Tensile Properties of VGCF Reinforced Carbon Composites, J. Mater. Sci.,
34, 229, 1999.
7. Ishioka, M., Okada, T., and Matsubara, K., Mechanical Properties of Vapor-grown
Carbon fibers Prepared from Benzene in Linz-Donawitz Converter Gas by Floating
Catalyst Method, J. Mater. Res., 7, 11, 3019, 1992.
8. Applied Sciences, Inc. ( http://www.apsci.com)
9. Uchida, T., Anderson, D.P., Minus, M., and Kumar, S., Morphology and Modulus
of Vapor Grown Carbon Nano Fiber, J. Mater. Sci., in press.
10. Tibbetts, G.G. and McHugh, J.J., Mechanical Properties of Vapor-grown Carbon
Fiber Composites with Thermoplastic Matrices, J. Mater. Res., 14, 7, 2871, 1999.
11. Lozano, K. and Barrera, E.V., Nanofiber-reinforced Thermoplastic Composites.
I. Thermoanalytical and Mechanical Analyses, J. Appl. Polym. Sci., 79, 1, 125,
2001.
12. Kuriger, R.J., Alam, M.K., Anderson, D.P., and Jacobsen, R.L., Processing and
Characterization of Aligned Vapor Grown Carbon Fiber Reinforced Polypropylene,
Composites A, 33, 53, 2002.
13. Caldeira, G., Maia, J.M., Carneiro, O.S., Covas, J.A., and Bernardo, C.A., Production
and Characterization of Innovative Carbon Fiber Polycarbonate Composites, Polym.
Composites, 19, 2, 147, 1998.
14. Carneiro, O.S. and Maia, J.M., Rheological Behavior of (Short) Carbon Fiber/
Thermoplastic Composites. Part II: The Influence of Matrix Type, Polym. Compos-
ites, 21, 6, 970, 2000.
15. Pogue, R.T., Ye, J., Klosterman, D.A., Glass, A.S., and Chartoff, R.P., Evaluating
Fiber-matrix Interaction in Polymer-matrix Composites by Inverse Gas Chromatogra-
phy, Composites A, 29, 1273, 1998.
16. Cooper, C.A., Ravich, D., Lips, D., Mayer, J., and Wagner, H.D., Distribution and
Alignment of Carbon Nanotubes and Nanofibrils in a Polymer Matrix, Composites
Sci. Technol., 62, 1105, 2002.
17. Sandler, J., Windle, A.H., Werner, P., Altstadt, V., Es, M.V., and Shaffer, M.S.P.,
Carbon Nanofibre-reinforced Poly(ether ether ketone) Fibres, J. Mater. Sci., 38,
10, 2135, 2003.
18. Cadek, M., Le Foulgoc, B., Coleman, J.N., Barron, V., Sandler, J., Shaffer, M.S.P.,
Fonseca, A., van Es., M., Schulte, K., and Blau, W.J., AIP Conference Proceed-
ings: Structural and Electronic Properties of Molecular Nanostructures, 633,
562, 2002.
19. Patton, R.D., Pittman, C.U., Wang, L., and Hill, J.R., Vapor Grown Carbon Fiber
Composites with Epoxy and Poly(phenylene sulfide) Matrices, Composites A, 30,
1081, 1999.
20. Uchida, T., Dang, T., Min, B.G., Zhang, X., and Kumar, S., Processing, Structure, and
Mechanical Properties of PBZT/carbon Nano Fiber Composite Fiber, Composites B,
in press.
21. Kumar, S., Doshi, H., Srinivasarao, M., Park, J.O., and Schiraldi, D.A., Fibers from
Polypropylene/Nano Carbon Fiber Composites, Polymer, 43, 1701, 2002.
22. Ma, H., Zeng, J., Realff, M.L., Kumar, S., and Schiraldi, D.A., Processing, Structure,
and Properties of Fibers from Polyester/Carbon Nanofiber Composites, Composites
Sci. Technol., 63, 1617, 2003.
23. Kozey, V.V., Jiang, H., Mehta, V.R., and Kumar, S., Compressive Behavior of
Materials. 2. High-performance Fibers, J. Mater. Res., 10, 4, 1044, 1995.
24. Zeng, J., Saltysiak, B., Johnson, W.S., Schiraldi, D.A., and Kumar, S., Processing
and Properties of Poly(methyl methacrylate)/Carbon Nano Fiber Composites, Com-
posites B, 35, 173, 2004.
25. Iijima, S., Helical Microtubules of Graphitic Carbon, Nature, 354, 56, 1991.
26. Baughman, R.H., Zakhidov, A.A., and de Heer, W.A., Carbon Nanotubesthe Route
toward Applications, Science, 297, 787, 2002.
27. Ajayan, P.M., Nanotubes from Carbon, Chem. Rev., 99, 1787, 1999.
28. Ajayan, P.M. and Zhou, O.Z., in Dresselhaus, M.S., Dresselhaus, G., and Avouris,
Ph., Eds., Carbon Nanotubes: Synthesis, Structure, Properties and Applications,
Springer-Verlag, Heidelberg, 2001, pp. 391425.
29. Thostenson, E.T., Ren, Z., and Chou, T.-W., Advances in the Science and Technology
of Carbon Nanotubes and their Composites: a Review, Composites Sci. Technol., 61,
1899, 2001.
30. Lau, K.-T. and Hui, D., The Revolutionary Creation of New Advanced Materials
Carbon Nanotube Composites, Composites B, 33, 263, 2002.
31. Kumar, S., Dang, T.D., Arnold, F.E., Bhattacharyya, A.R., Min, B.G., Zhang, X.,
Vaia, R.A., Park, C., Adams, A.A., Hauge, R.H., Smalley, R.E., Ramesh, S., and
Willis, P.A., Synthesis, Structure, and Properties of PBO/SWNT Composites, Mac-
romolecules, 35, 9039, 2002.
32. Sreekumar, T.V., Liu, T., Min, B.G., Guo, H., Kumar, S., Hauge, R.H., and Smalley,
R.E., Polyacrylonitrile Single-walled Carbon Nanotube Composite Fibers,
Advanced Mater., 16, 1, 58, 2004.
33. Dunn, M.L., Ledbetter, H.L., Heyliger, P.R., and Choi. C.S., Elastic Constants of
Textured Short-fiber Composite, J. Mech. Phys. Solids, 44, 1509, 1996.
34. Salvetat, J.-P., Briggs, G.A.D., Bonard, J.-M., Bacsa, R.R., Kulik, A.J., Stockli, T.,
Burnham, N.A., and Forro, L., Elastic and Shear Moduli of Single-walled Carbon
Nanotube Ropes, Phys. Rev. Letts., 82, 944, 1999.
35. Liu, T. and Kumar, S., Effect of Orientation on the Modulus of SWNT Films and
Fibers, Nano Lett., 3, 647, 2003.
36. Liu, T., Sreekumar, T.V., Kumar, S., Hauge, R.H., and Smalley, R.E., SWNT/PAN
Composite Film-based Supercapacitors, Carbon, 41, 2440, 2003.
37. Uchida, T. and Kumar, S., Single Wall Carbon Nanotube Dispersion and Exfoliation
in Polymers, J. Appl. Polym. Sci., in press.
38. Min, B.G., Sreekumar, T.V., Uchida, T., and Kumar, S., Oxidative Stabilization of
PAN/SWNT Composite Fiber, Carbon, 43, 599, 2005.
39. Sen, K., Bahrami, S.H., and Bajaj, P., High-performance Acrylic Fibers, J. Macro-
mol. Sci. Rev. Macromol. Chem. Phys., C36, 1, 1996.
40. Martin, S.C., Liggat, J.J., and Snape, C.E., In Situ NMR Investigation into the
Thermal Degradation and Stabilisation of PAN, Polym. Degrad. Stab., 74, 407, 2001.
41. Donnet, J.B., Wang, T.K., Rebouillat, S., and Peng, J.C.M., Carbon Fibers, Marcel
Dekker, New York, 1998.
42. Watt, W., Chemistry and Physics of the Conversion of Polyacrylonitrile Fibers into
High-Modulus Carbon Fibers, in Watt, W. and Perov, B.V., Eds., Strong Fibers, Vol.
1, North-Holland, Amsterdam, 1985, pp. 327387.
43. Johnson, W., The Structure of PAN Based Carbon Fibers and Relationship to Physical
Properties, in Watt, W. and Perov, B.V., Eds., Strong Fibers, Vol. 1, North-Holland,
Amsterdam, 1985, pp. 389443.
44. Kumar, S., Anderson, D.P., and Crasto, A.S., Carbon-fiber Compressive Strength
and its Dependence on Structure and Morphology, J. Mater. Sci., 28, 423, 1993.
45. Kumar, S., Adams, W.W., and Helminiak, T.E., Uniaxial Compressive Strength of
High Modulus Fibers for Composites, J. Reinforced Plastics Composites, 7, 108,
1988.
46. Guo, H., et al., unpublished data.
47. Zhang, X., Sreekumar, T.V., Liu, T., and Kumar, S., Gel Spinning of PVA/SWNT
Composite Fiber, Polymer, 45, 8801, 2004.
48. Zhang, X., Liu, T., Sreekumar, T.V., Kumar, S., Moore, V.C., Hauge, R.H., and Smalley,
R.E., Poly(vinyl alcohol)/SWNT Composite Film, Nano Lett., 3, 9, 1285, 2003.
49. Liu, J., Wang, T., Uchida, T., and Kumar, S., SWNT Core-polymer Shell Nano
Fiber, J. Appl. Polym. Sci., in press.
50. Dzenis, Y., Spinning Continuous Fibers for Nanotechnology Science, 304, 1917, 2004.
51. Yu, M., Files, B.S., Arepalli, S., and Ruoff, R.S., Tensile Loading of Ropes of Single
Wall Carbon Nanotubes and Their Mechanical Properties, Phys. Rev. Lett., 84, 24,
55525555, 2000.
52. Sreekumar, T.V., Liu, T., Kumar, S., Ericson, L.M., Hauge, R.H., and Smalley, R.E.,
Single-wall Carbon Nanotube Films, Chem. Mater., 15, 175, 2003.
53. Zhang, X., Sreekumar, T.V., Liu, T., and Kumar, S., Properties and Structure of Nitric
Acid Oxidized Single Wall Carbon Nanotube Films, J. Phys. Chem. B. 108, 16435, 2004.
54. Calvert, P., Nanotube Composites: A Recipe for Strength, Nature, 399, 210, 1999.
55. Holzinger, M., Steinmetz, J., Samaille, D., Glerup, M., Paillet, M., Bernier, P., Ley,
L., and Graupner, R., [2+1] Cycloaddition for Cross-linking SWCNTs, Carbon,
42, 944, 2004.
56. Kis, A., Csanyi, G., Salvetat, J.P., Lee, T.N., Couteau, E., Kulik, A.J., Benoit, W.,
Brugger, J., and Forro, L., Reinforcement of Single-walled Carbon Nanotube Bundles
by Intertube Bridging, Nature Mater., 3, 153, 2004.
57. Kuznetsova, A., Mawhinney, D.B., Naumenko, V., Yates, J.T., Jr., Liu, J., and Smalley,
R.E., Enhancement of Adsorption Inside of Single-walled Nanotubes: Opening the
Entry Ports Chem. Phys. Lett., 321, 292, 2000.
58. Yu, R., Chen, L., Liu, Q., Lin, J., Tan, K., Ng, S.C., Chan, H., Xu, G., and Andy
Hor, S.T., Platinum Deposition on Carbon Nanotubes via Chemical Modification.
Chem. Mater., 10, 718, 1998.
59. Sun, Y., Huang, W., Lin, Y., Fu, K., Kitaygorodskiy, A., Riddle, L.A., Yu, Y.J., and
Carroll, D.L., Soluble Dendron-functionalized Carbon Nanotubes: Preparation,
Characterization, and Properties, Chem. Mater., 13, 2864, 2001.
60. Hamon, M.A., Chen, J., Hu, H., Chen, Y., Itkis, M.E., Rao, A.M., Eklund, P.C., and
Haddon, R.C., Dissolution of Single-Walled Carbon Nanotubes, Adv. Mater., 11,
834, 1999.
61. Chen, J., Rao, A.M., Lyuksyutov, S., Itkis, M.E., Hamon, M.A., Hu, H., Cohn,
R.W., Eklund, P.C., Colbert, D.T., Smalley, R.E., and Haddon, R.C., Dissolution
of Full-length Single-walled Carbon Nanotubes, J. Phys. Chem. B, 105, 2525,
2001.
62. Jia, Z., Wang, Z., Liang, J., Wei, B., and Wu, D., Production of Short Multi-walled
Carbon Nanotubes, Carbon, 37, 903, 1999.
63. Mawhinney, D.B., Naumenko, V., Kuznetsova, A., and Yates, J.T., Infrared Spectral
Evidence for the Etching of Carbon Nanotubes: Ozone Oxidation at 298 K, J. Am.
Chem. Soc., 122, 2383, 2000.
64. Zhang, Y., Shi, Z., Gu, Z., and Iijima, S., Structure Modification of Single-wall
Carbon Nanotubes, Carbon, 38, 2055, 2000.
65. Zhang, M., Yudasaka, M., and Iijima, S., Diameter Enlargement of Single-wall
Carbon Nanotubes by Oxidation, J. Phys. Chem. B, 108, 149, 2004.
66. Hu, H., Zhao, B., Itkis, M.E., and Haddon, R.C., Nitric Acid Purification of Single-
walled Carbon Nanotubes, J. Phys. Chem. B, 107, 13838, 2003.
67. Bhattacharyya, A.R., Sreekumar, T.V., Liu, T., Kumar, S., Ericson, L.M., Hauge,
R.H., and Smalley, R.E., Crystallization and Orientation Studies in Polypropy-
lene/single Wall Carbon Nanotube Composite, Polymer, 44, 2373, 2003.
68. Dresselhaus, M.S. and Eklund, P.C., Phonons in Carbon Nanotubes, Advan. Phys.,
49, 705, 2000.
69. Pimenta, M.A., Marucci, A., Empedocles, S.A., Bawendi, M.B., Hanlon, E.B., Rao,
A.M., Eklund, P.C., Smalley, R.E., Dresselhaus, G., and Dresselhaus, M.S., Raman
Modes of Metallic Carbon Nanotube, Phys. Rev. B, 58, R16016, 1998.
70. Menna, E., Negra, F.D., Fontana, M.D., and Meneghetti, M., Selectivity of Chemical
Oxidation Attack of Single-wall Carbon Nanotubes in Solution, Phys. Rev. B, 68,
193412, 2003.
71. Brown, S.D.M., Jorio, A., Corio, P., Dresselhaus, M.S., Dresselhaus, G., Satio, R.,
and Kneipp, K., Origin of the Breit-Wigner-Fano Lineshape of the Tangential G-
band Feature of Metallic Carbon Nanotubes, Phys. Rev. B, 63, 1554141, 2001.
72. Gommans, H.H., Alldredge, J.W., Tashiro, H., Park, J., Magnuson, J., and Rinzler,
A.G., Fibers of Aligned Single-walled Carbon Nanotubes: Polarized Raman Spec-
troscopy, J. Appl. Phys., 88, 2509, 2000.
73. Anglaret, E., Righi, A., Sauvajol, J.L., Bernier, P., Vigolo, B., and Poulin, P., Raman
Resonance and Orientational Order in Fibers of Single-wall Carbon Nanotubes, Phys.
Rev. B, 65, 1654261, 2002.
74. Hwang, J., Gommans, H.H., Ugawa, A., Tashiro, H., Haggenmueller, R., Winey, K.I.,
Fischer, J.E., Tanner, D.B., and Rinzler, A., Polarized Spectroscopy of Aligned
Single-wall Carbon Nanotubes, Phys. Rev. B, 62, R13310, 2000.
75. Haggenmueller, R., Gommans, H.H., Rinzler, A.G., Fischer, J.E., and Winey, K.I.,
Aligned Single-wall Carbon Nanotubes in Composites by Melt Processing Methods,
Chem. Phys. Lett., 330, 219, 2000.
76. Fischer, J.E., Zhou, W., Vavro, J., Llaguno, M.C., Guthy, C., Haggenmueller, R.,
Casavant, M.J., Walters, D.E., and Smalley, R.E., Magnetically Aligned Single Wall
Carbon Nanotube Films: Preferred Orientation and Anisotropic Transport Properties,
J. Appl. Phys., 93, 2157, 2003.
77. Liu, T. and Kumar, S., Quantitative Characterization of SWNT Orientation by
Polarized Raman Spectroscopy, Chem. Phys. Lett., 378, 257, 2003.
13
Surface Patterning Using
Self-Assembled
Monolayers:
A Bottom-Up Approach
to the Fabrication
of Microdevices
Lakshmi Supriya and Richard O. Claus
CONTENTS
13.1 Introduction
13.2.1 Materials
13.2.2 Substrate Preparation
13.2.3 Deposition of SAM
13.2.4 Patterning
13.2.5 Deposition of Multilayers
13.3.1 Patterning on Silicon
13.3.2 Patterning on Polyethylene and Kapton
13.4 Conclusions and Applications

Problems
Acknowledgments
References
13.1 INTRODUCTION
The bottom-up approach to the fabrication of microscale devices requires the ability
to successfully pattern surfaces. This patterning can be achieved by the use of self-
assembled monolayers. In this chapter, surface patterning results with respect to
making photonic crystal devices and strain gauges are presented. Patterning is
achieved using two approaches. For the photonic crystal devices, ultraviolet (UV)
irradiation is used to pattern self-assembled monolayers, and for the strain gauge a
photoresist-based process is used. The electrostatic self-assembly method is used to
make patterned multilayers on silicon for the photonic crystal devices. Conducting
films of gold were deposited on flexible substrates for making strain gauges. The
films were deposited using a completely solution-based process. These approaches
are very cost-effective compared to the present microdevice fabricating technologies
discussed next, and the entire procedure can be done on a laboratory benchtop.
One of the most important fabrication technologies in the microelectronics indus-
try is the top-down approach to making microdevices, whether microelectromechan-
ical systems (MEMS), microfluidics, optical systems such as micromirrors, or com-
munication and information storage devices. In this approach, material is first laid
down, and it is removed gradually from top to bottom in patterns depending on the
application. The conventional microfabrication technologies have limited the choice
of materials; silicon is still the most widely used material of choice. This in turn
limits the applications of the devices. For example, it would not be possible to make
a biocompatible device for use in vivo with conventional materials used for micro-
electronics. The bottom-up approach to fabricating micro- and nanodevices has been
an area of considerable recent research.
12
The advantages of this approach over the
conventional top-down approach used in the microelectronics industry are many. It
offers considerable versatility in the materials used. Different types of organic mate-
rials that are more easily tailorable than ceramics, such as polymers, nanoparticles,
and self-assembled monolayers, are commonly studied. This approach also minimizes
waste of material and does not require sophisticated equipment, leading to much
lower costs and to molecular-scale control in device fabrication.
The key to such an approach is to successfully pattern the substrates and selec-
tively deposit materials forming three-dimensional structures. Patterning of sub-
strates using self-assembled monolayers (SAMs) is an area that has generated much
interest. Self-assembled monolayers
3
are long-chain alkanes that spontaneously
chemisorb onto the surfaces of appropriate materials. These layers offer control of
structure at the molecular level and can act as ultra-thin resists in lithographic
processes. Many problems associated with current patterning techniques, such as
the limit on the smallest feature size, optical diffraction effects, depth of focus,
shadowing, and undercutting can be minimized. Self-assembly leads to equilibrium
structures that are at or near the thermodynamic minimum, and as a result these
systems tend to be self-healing and defect-rejecting.
In order to fabricate functional microdevices, it is essential to form three-dimen-
sional structures on the self-assembled monolayer. These structures can be built if
multilayers can be grown in patterns of the desired structure. There are a number of
patterning techniques for SAMs that have been studied: soft lithography techniques
such as microcontact printing,
4
replica molding,
5
microtransfer molding,
6
selective
transformation of terminal functionalities by either photochemical
710
or chemical
1113
methods, selective removal of monolayer film by irradiation
1417
and micromachin-
ing,
18
and selective deposition of polyelectrolytes
19
and nanoparticles. A more recent
method of patterning is the use of scanning probe microscopy tips, which has dras-
tically reduced the sizes of the patterns formed and is so effective that it can be used
to manipulate single molecules.
20
Techniques such as nanografting
21
and dip pen
nanolithography
22
use the tip of an AFM probe to selectively remove and deposit
molecules onto the surface. Scanning tunneling microscopy probes have also been
used to pattern SAMs using electrochemical methods by applying a bias between the
conductive probe and the substrate.
23
In this chapter, results with regard to the fabrication of different types of devices
using patterned SAMs using the bottom-up approach are presented. A brief descrip-
tion of the type of prototype devices investigated follows. A rst set of representative
experiments on silicon substrates was aimed at fabricating spatially periodic photonic
crystals. Photonic crystals are a class of periodic composite structures consisting of
spatially varying low and high refractive index materials that exhibit a forbidden
band, or photonic band gap, of frequencies wherein incident electromagnetic waves
cannot propagate.
24
Introducing defects into the ordinarily symmetrical structure
disrupts the periodicity of the crystal-dening microcavities, and in some cases
waveguides, where incident electromagnetic energy can become localized or allowed
to propagate. The creation of patterns on silicon and formation of multilayers using
the electrostatic self-assembly process
25
is a possible method of fabricating these
devices.
26
Microstrain gauges are other spatially patterned devices that may be fab-
ricated in situ using self-assembly. These can be achieved by patterning conducting
lms on exible substrates. The basic theory behind these devices is that a change
in the length and geometry of an electrical conductor causes a change in its resistance,
and this change can be quantied and used to infer elongation. Using self-assembly,
patterned gauges may be formed by depositing conducting gold on substrates such
as polyethylene and measuring the change in resistance of the conductor versus strain.
13.2.1 MATERIALS
Three types of substrates for patterning were studied: silicon (111); Kapton
(DuPont), a polyimide; and polyethylene. The SAMs used were octadecyltrichlo-
rosilane (OTS) (Aldrich), heptadecauoro 1, 1, 2, 2, tetrahydrodecyl trichlorosilane
CF
3
(CF
2
)
7
CH
2
CH
2
SiCl
3
(CF) (Gelest), 3-aminopropyltrimethoxy silane (APS)
(Gelest) and 3-mercaptopropyltrimethoxy silane (MPS) (Gelest). The polyelectrolyte
used was sodium polystyrene sulfonate (SPS), and the nanoparticles used were
silicon dioxide (Nissan Chemical Industries), titanium dioxide, and gold, which were
synthesized using a slight modication of previously described methods.
27
13.2.2 SUBSTRATE PREPARATION
Silicon wafers, cut into small 2-cm squares, were rinsed thoroughly with dichlo-
romethane and methanol and dried under a stream of nitrogen. The samples were
treated with a 7: 3 v/v mixture of H
2
SO
4
/H
2
O
2
for 20 min at 80C (piranha treatment)
and rinsed thoroughly with nanopure water (18 M) from a Barnstead nanopure water
system. The Kapton and polyethylene were plasma treated using a March Plasmod
plasma etcher in an argon atmosphere for ~2 min at 0.20.4 torr and 50 W power.
13.2.3 DEPOSITION OF SAM
OTS SAM was deposited in a glove bag under nitrogen atmosphere. The piranha-
treated silicon substrates were immersed in a 1% (v/v) solution of OTS in dicyclo-
hexyl (Aldrich) for 1 h with stirring. After the reaction was over, the substrates
were rinsed thoroughly with dichloromethane. CF SAM was also deposited in a
similar manner, with Isopar-G (Exxon) acting as the solvent in this case. The entire
deposition process was carried out in a nitrogen atmosphere and dry conditions.
The deposition of APS and MPS was carried out in ambient conditions. The
substrates were immersed in a 1% (v/v) solution of APS/MPS for 15 min with
stirring. After that they were rinsed thoroughly in methanol and heated at 110qC
for 1
1
/
2
h and then dried under vacuum at 50qC. All the SAM depositions were
carried out in less than an hour after the substrate preparation.
13.2.4 PATTERNING
Patterning of the SAMs was carried out by irradiation with an Hg(Ar) lamp (Model
6035, Oriel Instruments) at a wavelength of 184 nm. The OTS SAMs were photo-
lyzed for 1 h, while all the others were photolyzed for 3 h. The masks consisted
of meshes used for transmission electron microscopy, and masks were also made
on aluminum sheets when larger scale patterns were required. Patterning on the
flexible substrates was carried out by using a positive photoresist S1813 (Shipley).
The photoresist was spin-coated onto the silane-coated substrates, and then irradi-
ated with broadband UV (350450 nm) (Oriel Instruments Model 92530) through
a mask, developed using the Microposit 351 developer. Then gold was coated onto
the entire substrate and the photoresist was removed using acetone. At that time
the gold coated over the photoresist was also removed, leaving a pattern of con-
ducting lines.
13.2.5 DEPOSITION OF MULTILAYERS
Multilayers were deposited on silicon using electrostatic self-assembly. TiO
2
and
SiO
2
were alternately deposited on patterned OTS, and TiO
2
and SPS on patterned
CF. The patterned samples were first dipped in the positively charged solution for
3 min, sonicated in nanopure water for 2 min, dipped in the oppositely charged
solution for 3 min and sonicated again, and this cycle was repeated to obtain the
desired number of multilayers. On the polymer substrates, gold was deposited using
previously described methods.
28
13.3.1 PATTERNING ON SILICON
The piranha treatment cleans the surface exposing the hydroxyl groups on the
surface of the silicon. The deposition of silanes on these surfaces occurs by the
hydrolysis of the alkoxy or trichloro groups on the silane and subsequent reaction
with the surface hydroxyl groups.
3
The deposition of OTS and CF was characterized
by contact angle measurements. After the piranha treatment the silicon substrate
was completely wetted with contact angles as low as ~10. The deposition of OTS
or CF raised the contact angles. For OTS the angles were between 90 and 115,
and for CF the contact angles were slightly higher, around 120. This was expected
because CF has a much lower surface energy than OTS due to the presence of
fluorine groups. After irradiation with UV light at 184 nm, both the SAMs were
degraded and removed from the substrate.
17
This is evidenced by the drop in contact
angles to ~10 upon photolysis. The CF took a longer time than OTS to photolyze
completely and this can be attributed to the strong CF bonds. Figure 13.1 illustrates
the photolysis process on OTS.
On OTS and CF, the multilayers were formed by electrostatic self-assembly.
Electrostatic self-assembly involves dipping a charged substrate alternately into
solutions of an anionic charged species and a cationic charged species; the electro-
static force between oppositely charged molecules acts as the binding force holding
the molecules together. Patterning on OTS and CF creates regions of charge and
regions where there is no charge and the charged molecules become selectively
adsorbed on the charged regions. TiO
2
was the positively charged nanoparticle for
OTS and CF; SiO
2
was the negatively charged nanoparticle for the patterned OTS,
and SPS was the negatively charged polyelectrolyte used for CF. After photolysis,
the negatively charged substrate is exposed and the charged particles selectively
adsorb on the charged regions and are bound by electrostatic interactions. Figure
13.2 shows SEM images of the patterns formed on OTS- and CF-coated substrates.
The pattern contained 7.5 u 7.5 Pm square holes with a center-to-center spacing of
12.5 Pm. The lengths of all the scale bars in the figure were 10 Pm. The dark regions
are the regions of deposition in all the images except in Fig. 13.2(a), where the
opposite applies. This is probably due to some differential charging effect when
imaging the samples.
The deposition of the polyelectrolyte, SPS, was successful for the CF, while
good patterns were not obtained for the OTS. This can be explained by the fact that
the negatively charged groups on the polyelectrolyte are repelled by the electroneg-
ative fluorine atom on the CF and are thus selectively deposited on the regions where
FIGURE 13.1 Conceptual illustration of the patterning process for OTS.
O
O
O
O
O O O
O O O
O
O O
O O
O
O O
UV
Si
Si
CH
3
(CH
3
)
17
(CH
3
)
17
(CH
3
)
17
(CH
3
)
17 (CH
3
)
17
(CH
3
)
17
CH
3
CH
3
CH
3
CH
3
CH
3
Hydrophilic
Hydrophobic
Si Si Si Si
Si Si Si Si
OH OH
Mask
CF is absent. On the OTS SAM there was no such interaction and SPS, being a large
molecule, may be attached on the OTS regions also due to some hydrophobic inter-
actions. These selectively grown ESA nanocomposites represent two-dimensional
photonic crystal structures composed of high refractive index pillars suspended in air.
Measurements on a Rudolph Auto EL ellipsometer revealed that the TiO
2
-SiO
2
cor-
responded to high index pillars with an average refractive index of 1.6.
13.3.2 PATTERNING ON POLYETHYLENE AND KAPTON
Polyethylene and Kapton are flexible polymers suitable for making patterned elec-
trodes to fabricate microstrain gauges in situ. This requires selective deposition of
conductive layers, and gold was the material of choice because of the ease of
deposition of a conductive film using a completely solution-based process. The
resistance of a conductor is dependent on its length and cross-sectional geometry,
and a change in the geometry would lead to a change in the resistance and this can
be measured quantitatively. This principle is used in strain gauges also, where a
change in the length will cause a change in the resistance and the strain can be
measured. In order to increase the device sensitivity, the length of the conductor
should be increased and this can be done using geometry as shown in Fig. 13.3.
The surface of the polymers is plasma treated to deposit the silanes. Plasma
treatment of the polymers creates free radicals on the surface, which when exposed
to air form oxygenated species.
29
These can then react with the silanes to form a
monolayer on the surface. Table 13.1 shows the contact angles for substrates after
FIGURE 13.2 SEM images of patterned multilayers on silicon (a) 20, (b) 30 TiO
2
-SiO
2
bilayers on OTS-treated silicon, (c) 50 TiO
2
SPS bilayers on CF-treated silicon.
plasma treatment and SAM deposition. Figure 13.4 shows the x-ray photoelectron
spectroscopy (XPS) data corresponding to the different treatments on Kapton. A
comparison of the untreated and the plasma-treated surfaces shows an increase in the
oxygen concentration, as expected. The deposition of the SAM on the substrate gives
rise to a polysiloxane backbone. Figure 13.4(c) shows the siloxane peak at 102.4 eV,
which confirms the formation of the SAM. Similar results were also obtained for
polyethylene.
The deposition of gold on the SAMs takes advantage of the fact that gold has
a large affinity for groups such as amino, mercapto, and cyano compounds.
27
The
choice of the SAMs was based on this, and amino- and mercapto-terminated SAMs
were used in our experiments. The formation of conductive layers of gold nanopar-
ticles is done by a seeding method as described in a previous study.
28
In brief, the
SAM-coated substrates are dipped in colloidal gold solution for 23 h and gold
particles are attached to the substrate. Then more gold is reductively deposited on
the gold already attached, leading to complete coverage. Figure 13.5 shows the SEM
images of the formation of complete coverage gold on Kapton. All scale bars are
100 nm.
FIGURE 13.3 (a) An illustration of the patterned geometry of a strain gauge (From
http://www.vishay.com/brands/measurements_group/strain_gages/mm.htm). (b) Patterned
and printed strain gauge.
TABLE 13.1
Contact Angles for Polymers after Plasma Treatment and SAM Deposition
Kapton Polyethylene
Untreated 71q 94q
Plasma treated 18q 41q
After SAM deposition 84q 95q
FIGURE 13.4 XPS spectra of (a) untreated (b) plasma treated, and (c) Si peak of APS-
deposited Kapton. (Reprinted with permission from Langmuir, 2004, 20, 88708876. Copy-
right 2004, American Chemical Society.)
10
9
8
7
6
5
4
3
2
1
0
1000 800 600 400 200 0
C

K
L
L
0

K
V
V
0
1
s
0
2
s
N
1
s
< C1s
Binding energy, eV
(a)
10
9
8
7
6
5
4
3
2
1
0
1000 800 600 400 200 0
Binding energy, eV
(b)
C

K
L
L
0

K
V
V
0
1
s
0
2
s
S
i
2
s
S
i
2
p
N
1
s
C
1
s
10
9
8
7
6
5
4
3
2
1
0
114 112 110 108 106 104 102 100 98 96
Binding energy, eV
(c)
102.4
The resistance of the samples was measured using a four-point probe. The
resistivities on both polyethylene and Kapton were about 1 ohm/sq. The thickness of
a layer of gold was found to be ~300 nm using cross-section SEM. Table 13.2 depicts
the strain resistivity data for gold on Kapton, and as expected the resistance is found
to increase as a function of length.
The patterning on these substrates was done to generate conductive lines, as
shown in Fig. 13.3. Patterning was done using a method developed by Hua et al.
30
Figure 13.6 depicts the patterns generated on Kapton, which consist of alternate
TABLE 13.2
Strain-Resistivity Data for Gold on Kapton
Load (N) Resistivity 10
4
(cm) Strain
0 0.95 0
10 1.28 0.078
20 1.26 0.17
70 3.2 1.875
FIGURE 13.5 SEM images of (a) gold monolayer, (b) one-layer seeded gold, and (c) two-
layer seeded gold on Kapton. (Reprinted with permission from Langmuir, 2004, 20,
88708876. Copyright 2004, American Chemical Society.)
FIGURE 13.6 Patterned gold on Kapton.
conducting lines. The size of the lines is about 1 mm. Although the patterns shown
here are large, with masks of smaller size, patterns of 1 Pm or smaller can also be
made. The conductivity of the film was not affected by the patterning process and
was the same as for the unpatterned substrates.
13.4 CONCLUSIONS AND APPLICATIONS
It is possible to selectively deposit and create three-dimensional structures on silicon
using a combination of patterned self-assembled monolayers and electrostatic self-
assembly. The size of the smallest pattern created was 7.5 Pm, which was the smallest
mask utilized. However, with appropriate masks the pattern size can be reduced even
further. It was also possible to deposit conducting layers of gold on polymer sub-
strates such as polyethylene and Kapton for the fabrication of strain gauges. Pat-
terning was also successfully achieved to fabricate conducting lines. These results
show that fabrication of microdevices can be achieved by methods that are more
economical and easier than conventional microfabrication technologies that use
sophisticated equipment and involve increased costs. This method is so simple that
it can be performed on a laboratory benchtop. Future efforts are being directed
toward decreasing the sizes of patterns and using a maskless approach to patterning.
PROBLEMS
1. List the important trade-offs when considering bottom-up and top-down
fabrication technologies.
2. Perform a literature search on the subject of patterning of substrates using
self-assembled monolayers and list any new advances in techniques that
have been made relative to the techniques cited in this chapter.
3. Think of some new possible applications for surface patterning using the
SAM technique and list potential benefits.
ACKNOWLEDGMENTS
The authors gratefully acknowledge financial support from NASA and the U.S.
Army Research Laboratory and U.S. Army Research Office under contract/grant
number DAAD19-02-1-0275, Macromolecular Architecture for Performance (MAP)
MURI. We also thank Steve McCartney and the Materials Research Institute, Vir-
ginia Tech, for the SEM images, and Frank Cromer and the Department of Chemistry,
Virginia Tech, for the XPS data.
REFERENCES
1. Faul Charl, F.J. and Antonietti, M., Ionic self assembly: facile synthesis of supramo-
lecular materials Adv. Mat., 15, 9, 673683, 2003.
2. Zhao, X.-M., Xia, Y., and Whitesides, G.M., Soft lithography methods for nano-
fabrication J. Mater. Chem., 7, 7, 10691074, 1997.
3. Ulman, A., Formation and structure of self-assembled monolayers, Chem. Rev., 96,
15331554, 1996.
4. Kumar, A. and Whitesides, G.M., Features of gold having micrometer to centimeter
dimensions can be formed through a combination of stamping with an elastomeric
stamp and an alkanethiol ink followed by chemical etching, Appl. Phys. Lett., 63,
20022004, 1993.
5. Xia, Y., Kim, E., Zhao, X.-M., Rogers, J.A., Prentiss, M., and Whitesides, G.M.,
Complex optical surfaces formed by replica molding against elastomeric masters,
Science, 273, 347349, 1996.
6. Zhao, X.-M., Xia, Y., and Whitesides, G.M., Fabrication of three-dimensional micro-
structures: microtransfer molding, Adv. Mat., 8, 837840, 1996.
7. Bhatia, S.K., Hickman, J.J., and Ligler, F.S., New approach to producing patterned
biomolecular assemblies, J. Am. Chem. Soc., 114, 44324433, 1992.
8. Collins, R.J., Shin, H., DeGuire, M.R., Heuer, A.H., and Sukenik, C.H., Low tem-
perature deposition of patterned TiO
2
thin films using photopatterned self-assembled
monolayers, Appl. Phys. Lett., 69, 6, 860862, 1996.
9. Brandow, S.L., Chen, M.-S., Aggarwal, R., Dulcey, C.S., Calvert, J.M., and Dressick,
W.J., Fabrication of patterned amine reactivity templates using 4-chloromethylphe-
nylsiloxane self-assembled monolayer films, Langmuir, 15, 54295432, 1999.
10. Rozsnvai, L.F. and Wrighton, M.S., Selective electrochemical deposition of polya-
niline via photopatterning of a monolayer-modified substrate, J. Am. Chem. Soc., 116,
59935994, 1994.
11. Wasserman, S.R., Tao, Y.T., and Whitesides, G.M., Structure and reactivity of alkyl-
siloxane monolayers formed by reaction of alkyltrichlorosilanes on silicon substrates,
Langmuir, 5, 10741087, 1989.
12. Balachander, N. and Sukenik, C.N., Monolayer transformation by nucleophilic trans-
formation: applications to the creation of new monolayer assemblies, Langmuir, 6,
16211627, 1990.
13. Netzer, L. and Sagiv, J., A new approach to construction of artificial monolayer
assemblies, J. Am. Chem. Soc., 105, 674676, 1983.
14. Tarlov, M.J., Burgess, D.R.F., and Gillen, G., UV photopatterning of alkanethiolate
monolayers self-assembled on gold and silver, J. Am. Chem. Soc., 115, 53055306,
1993.
15. Dulcey, C.S., Georger, J.H., Jr., Krauthamer, V., Stenger, D.A., Fare, T.L., and Calvert,
J.M., Deep UV photochemistry of chemisorbed monolayers: patterned coplanar
molecular assemblies, Science, 252, 551554, 1991.
16. Dressick, W.J. and Calvert, J.M., Patterning of self-assembled films using lithographic
exposure tools, Jpn. J. Appl. Phys., 32, 1, 12B, 58295839, 1993.
17. Masuda, Y., Seo, W.S., and Koumoto, K., Selective deposition and micropatterning
of titanium dioxide on self-assembled monolayers from a gas phase, Langmuir, 17,
48764880, 2001.
18. Abbott, N.L., Folkers, J.P., and Whitesides, G.M., Manipulation of the wettability of
surfaces on the 0.1- to 1-micrometer scale through micromachining and molecular
self-assembly, Science, 257, 13801382, 1992.
19. Clark, S.L., Montague, M., Hammond, P.T., Selective deposition in multilayer assem-
bly: SAMs as molecular templates, Supramol. Sci., 4, 12, 141146, 1997.
20. Grandbois, M., Beyer, M., Rief, M., Clausen-Schaumann, H., and Gaubi, H.E., How
strong is a covalent bond? Science, 283, 17271730, 1999.
21. Xu, S. and Liu, G., Nanometer-scale fabrication by simultaneous nanoshaving and
molecular self-assembly, Langmuir, 13, 127129, 1997.
22. Piner, R.D., Zhu, J., Xu, F., Hong, S., and Mirkin, C.A., Dip-pen nanolithography,
Science, 283, 661663, 1999.
23. Schoer, J.K., Zamborini, F.P., and Crooks, R.M., Scanning probe lithography. 3.
Nanometer-scale electrochemical patterning of Au and organic resists in the absence
of intentionally added solvents or electrolytes, J. Phys. Chem., 100, 1108611091,
1996.
24. Joannopoulos, J.D., Meade, R.M., and Winn, J.N., Photonic Crystals: Molding the
Flow of Light, Princeton University Press, Princeton, NJ, 1995.
25. Decher, G. and Hong, J.D., Buildup of ultrathin multilayer films by a self-assembly
process. 1. Consecutive adsorption of anionic and cationic bipolar ampiphiles on
charged surfaces, Makromol. Chem. Macromol. Symp., 46, 321327, 1991.
26. Huie, K.C., Chandran, A., Claus, R.O., Nelson, R., and Supriya, L., Photonic crystal
fabrication using electrostatic self-assembly photonics, paper presented at 6th Annual
Conference on Optoelectronics, Fiber-Optics, and Photonics, December 1618, 2002,
TIFR, Mumbai.
27. Grabar, K.A., Freeman, R.G., Hommer, M.B., and Natan, M.J., Preparation and
characterization of Au colloid monolayers, Anal. Chem., 67, 735743, 1995.
28. Brown, K.R., Lyon, L.A., Fox, A.P., Reiss, B.D., and Natan, M.J., Hydroxylamine
seeding of colloidal Au nanoparticles. 3. Controlled formation of conductive Au films,
Chem. Mater., 12, 314323, 2000.
29. Inagaki, N., Tasaka, S., and Baba, T., Surface modification of polyimide film surface
by silane coupling reactions for copper metallization, J. Adhesion Sc. Tech., 15, 7,
749762, 2001.
30. Hua, F., Cui, T., and Lvov, Y., Lithographic approach to pattern self-assembled multilay-
ers, Langmuir, 18, 67126715, 2002.
14
Enhancement of the
Mechanical Strength
of Polymer-Based
Composites Using
Carbon Nanotubes
Kin-Tak Lau, Jagannathan Sankar, and David Hui
CONTENTS
14.1 Introduction
14.2 Properties of Carbon Nanotubes
14.2.1 Experimental Measurements
14.2.2 Theoretical Study and Molecular Dynamics Simulation
14.2.3 Finite Element Modeling
14.3 Fabrication Processes of Nanotube/Polymer Composites
14.4 Interfacial Bonding Properties of Nanotube/Polymer Composites
14.4.1 Experimental Investigation
14.4.2 Theoretical Study and Molecular Dynamics Simulation
Problems
Acknowledgments
References
14.1 INTRODUCTION
Since the discovery of carbon nanotubes a decade ago,
1
many important studies
and results related to these nanostructural materials in different scientific and
engineering fields have emerged. The extraordinary mechanical, electrical, and ther-
mal properties of the nanotubes are governed by their atomic architecture, commonly
called their chiral arrangement. Ideally all carbon atoms in the nanotubes are
covalently bonded and form repeated close-packed hexagonal structures in each
layer or shell. Due to these chemically formed atomic arrangements, the carbon
nanotubes possess superior mechanical properties, and the nanotubes are stronger
than any known metallic materials. Many critical results have been reported recently
by using nanotubes as atomic force microscope (AFM) probes, conductive devices
in artificial muscles, and nanothermometers, and to store hydrogen for fuel cells.
25
In the United States, the investment in the development of fuel cells by storing
hydrogen atoms inside the cavities of nanotubes to supply electricity to microelec-
tromechanical (MEM) or even nanoelectromechanical (NEM) devices has been
increasing. Although all this work is still at the research stage, it demonstrates a
high potential for the development of nanotube-related products and components
for real-world applications.
In the past few years, many researchers and engineers from the advanced
composites community have attempted to use these tiny structural materials to
enhance the properties of conventional advanced composite structures by altering
their mechanical, thermal, and electrostatic behaviors for space and infrastructure
applications. However, in order to achieve these goals, more work is needed for
(1) understanding the mechanical properties of both single-walled and multi-
walled nanotubes, (2) investigating appropriate fabricating processes of nano-
tube/polymer composites, (3) clarifying the interfacial bonding properties between
the nanotubes and surrounding matrices, and (4) justifying the benefit based on the
strength improvement of composites after mixing with the nanotubes. It is under-
standable that research related to the aforementioned issues still has a long way to
go due to many uncertainties related to nanotube properties and their structural
integrity in nanotube/polymer composites.
In this chapter, a critical review of the aforementioned aspects is based on recent
research by the authors and other researchers. All these aspects cannot be considered
individually in the development of nanotube/polymer composites. Therefore, a dis-
cussion of how these aspects affect each other, as well as a detailed discussion of
each one, is given. Since this chapter is mainly focused on the mechanical properties
of nanotube/polymer composites, the fundamental physics of nanotubes is briefly
explained here.
14.2 PROPERTIES OF CARBON NANOTUBES
14.2.1 EXPERIMENTAL MEASUREMENTS
A carbon nanotube is similar to a flat graphene layer rolled up to form a tube where
both ends of the tube are sealed with semihemisphere caps. All the carbon atoms are
tightly bonded to each other to form a close-packed hexagonal structure in the form
of a circular tube, as shown in Fig. 14.1. If more than one layer of graphene layers
are rolled together to form a coaxial tube, this nanotube is called a multi-walled
nanotube (MWNT). The mechanical and electrical properties of nanotubes are gov-
erned by their atomic structures. In the past few years, many researchers have
attempted to measure the mechanical properties of nanotubes. However, due to their
small size, it is difficult to directly measure the strength of the nanotubes by using
traditional testing methods. Yu et al.
6
mounted MWNTs on two AFM tips and
conducted a tensile test. The whole stretching process was captured inside a scanning
electron microscope (SEM) as shown in Fig. 14.2. Since different sizes and types
of multi-walled nanotubes were tested, the results were quite diverse. The Youngs
moduli of the nanotubes ranged from 0.32 TPa to 1.47 TPa. In strength testing, almost
all the outer layers of the nanotubes broke first, and then pulling out of inner layers
occurred. According to the results in the literature, the maximum stress applied to the
nanotubes occurred in the outer layers. The larger diameter nanotubes could withstand
a larger tensile force. This conclusion may not coincide with some estimates of
strength from continuum models and molecular dynamics (MD) simulations.
An indirect tensile test on nanotubes by Demczyk et al.
7
demonstrated that the
tensile strength of MWNTs was about 0.15 TPa, which is far below theoretical
estimates and the known properties of a graphene layer. A telescoping action was
seen in some fractured nanotubes. This finding agreed with the work done in
FIGURE 14.1 Examples of carbon nanotubes.
FIGURE 14.2 SEM images of a nanotube linked between two opposing AFM tips before
tensile loading.
6
Two-dimensional
graphene sheet
Chiral tubule (4,3)
Armchair tubule (4,4)
Zigzag tubule (7,0)
Seamless nanotube
a
1
a
2
O
C
C
h

Reference 8, wherein a weak van der Waals attractive force between individual shells
of the MWNTs resulted in generating an extremely low frictional force among the
shells, and all the inner shells slid freely in their longitudinal direction (Fig. 14.3).
The principal result is that all the inner shells cannot contribute any strength to the
nanotubes while they are subjected to tension. Qi et al.
9
also measured the Youngs
moduli of MWNTs using a nanoindentation method. An indentor was pressed onto
vertical aligned nanotubes and the bending stiffness of the nanotubes was measured
and also calculated using classical bending theory. The effective bending and axial
elastic moduli ranged 0.91 a 1.24 TPa and 0.90 a 1.23 TPa, respectively. Since the
ineffectiveness of the stress transfer among different shells of the nanotubes was not
considered in the calculation, particularly in the bending case, the moduli estimated
by this method do not represent the true tensile moduli of the nanotubes.
14.2.2 THEORETICAL STUDY AND MOLECULAR
DYNAMICS SIMULATION
Because the sizes of nanotubes are extremely small, direct measurement of the
mechanical properties of nanotubes is extremely difficult. In the past few years,
much effort has been spent on using MD simulations to predict the properties of
nanotubes. Since the accuracy of the simulated results is highly dependent on the
size of the models as well as the capacity (memory size) and running speed of
computers, the results from different papers and approaches are therefore different.
Tu and Ou-Yang
10
have used the local density approximation method associated with
elastic shell theory to estimate the mechanical properties of single-walled nanotubes
(SWNTs) and MWNTs. They found that the Youngs modulus of the multi-walled
nanotubes decreased with an increase in the number of shell layers. The Youngs
modulus of the nanotubes can be determined by the following equation:
(14.1)
where E
m
and E
s
represent the Youngs moduli of MWNTs and SWNTs, respectively,
and n, t, and d denote the number of shell layers, average thickness of the shell
(|0.75 ), and spacing between the shells (|0.34 ), respectively. They have also
reduced Equation 14.1 to the continuum limit form, and the classic shell theory can
be used to describe the deformation of the nanotubes. Lau et al.
11
used the Tersoff-
Brenner bond order potential to represent the interaction between carbon atoms to
FIGURE 14.3 A telescoping action in a multi-walled nanotube.
7
E
n
n t d
t
d
E
m s
+

1 /
study the stretching motion of MWNTs. In their study, it was found that the outer
shells of the nanotubes took all the applied loads when the nanotubes were subjected
to tensile and torsion motions (Fig. 14.4). Since only a weak van der Waals inter-
action exists between the shells, external motions of the outer shell cannot be easily
transferred to inner shells. Eventually fracture at the outer shell was initiated
(Fig. 14.5) and this phenomena was similar to that observed in Reference 7, where
FIGURE 14.4 (Color figure follows p. 12.) MD-simulated results of three-layer MWNTs.
The inner two layers would not be affected by the surface layer when the MWNTs are
subjected to tensile (top) and torsional (bottom) loads.
FIGURE 14.5 Outer shell of MWNT pulling out of a matrix.
pullout of nanotubes was observed. Lau
12
has also pointed out that the size and
Youngs moduli of nanotubes are highly dependent on their chiral arrangement. The
diameter of a zigzag nanotube is generally smaller than that of an armchair type.
The radius of the first layer (an inner layer) of nanotubes can be determined by
using the rolled graphene sheet model:
(14.2)
where U
o
, m, n, and a
o
are the nonrelaxed radius and indices of the SWNTs and CC
bond distance (1.42 ), respectively.
The MD simulation has been popularly adopted in recent years to predict
the properties of nanostructural materials at the atomic scale with high accuracy.
At the early stage, several studies have used empirical force potential molecular
dynamic simulations to estimate the Youngs moduli of nanotubes and found that
their value was four times that of the diamond. However, those calculations were
based on the SWNTs of several angstroms in radius. To determine the mechanical
properties of nanotubes, the details of their atomic arrangements have to be
clearly understood. In general, there are three types of nanotube structures: zigzag
(n, 0), armchair (n, n), and chiral (n, m where n z m). Lau and Hui
13
wrote a
comprehensive review on the structures of nanotubes. To investigate the mechan-
ical properties of materials at the atomic scale by using MD simulations, the
interactions between neighboring atoms must be accurately calculated.
Two common approaches based on quantum mechanics and molecular mechan-
ics are used to simulate these interactions. Both approaches attempt to capture the
variation of system energy associated with the change in atomic positions by fol-
lowing Newtons second law, F ma. In carbon nanotubes, the mutual interactions
are described by force potentials from both bonding and nonbonding interactions.
The nonbonding interactions are either due to the van der Waals force (which can
be attractive or repulsive, depending on the distance between atoms), or to electro-
static interactions. The van der Waals force F
VDW
is most often modeled using the
Lennard-Jones potential function,
14
originally derived for inert gases. The general
form of this potential is
(14.3)
For van der Waals forces arising from dipole-dipole interactions, the attractive
part corresponds to m 6. The most common form of this potential is the so-
called (612) form:
(14.4)
o o
m n mn
a
+ +
2 2
3
r
r r
n
n
m
m
( )

r
r r
( )
j
(
,
\
,
(
j
(
,
\
,
(
,
,
,
]
]
]
]
4
12 6

The minimum of )(r) is determined by equating to zero the first-order derivative
of )(r) versus r. The van der Waals force between two carbon atoms can be estimated
from
(14.5)
The two parameters V and H can be estimated from experimental data such as the
equilibrium bond length (lattice parameters at equilibrium), equilibrium bond energy
(cohesive energy), and bulk modulus at equilibrium. The bonding energy (E
bond
) is
the sum of four different interactions among atoms, namely bond stretching (U
U
),
angle variation (U
T
), inversion (U
Z
), and torsion (U
W
),
15
written as
E
bond
= U
U
+ U
T
+ U
Z
+ U
W
(14.6)
A schematic illustration of each energy term and corresponding bond structure
for a graphene cell is shown in Fig. 14.6. The most commonly used functional
forms are
(14.7ad)
where dR
i
is the elongation of the bond identified by the label i, K
i
is the force
constant associated with the stretching of the i bond, and dT
j
and dZ
k
are the variance
of bond angle j and inversion angle k, respectively. C
j
and B
k
are force constants
associated with angle variance and inversion, respectively. A
i
is the barrier height to
rotation of the bond i; n
i
is the multiplicity that gives the number of minimums as
the bond is rotated through 2S.
15
To determine the tensile modulus of a SWNT subjected to uniaxial loading, it
is useful to observe that at small strains the torsion, the inversion, the van der Waals,
and the electrostatic interaction energy terms are small compared with the bond
F
d
dr r r r
VDW

j
(
,
\
,
(
j
(
,
\
,
(
,
,
,
]
]
]
24
2
12 6
]]
U K dR
i
i
i
( )
1
2
2
U C d
j
i
j

( )
1
2
2
U B d
k
k
k
( )
1
2
2
U A n
i
i
i i i
+ ( ) ,
]
]
1
2
1 cos
stretching and the angle variation terms. Thus the total energy of the single-wall
nanotube can be reduced to
(14.8)
The force constants K
i
and C
i
can be obtained from quantum mechanics (ab initio).
The average macroscopic elastic modulus and Poissons ratio were estimated to be
about 1.347 TPa and 0.261, respectively.
15
Such calculations may be performed using
either the force or the energy approach, by measuring the mechanical forces developed
between carbon atoms in nanotubes with different chiral arrangements. Lu
16
has used
the empirical force potential molecular dynamic simulation to investigate the properties
of nanotubes. The structure of the nanotubes was obtained by the conformational
mapping of a graphene sheet onto a cylindrical surface. The nanotube radius was
estimated by Equation 14.2. The average estimated tensile modulus of SWNTs and
MWNTs was about 1 TPa. The elastic properties were the same for all nanotubes with
a radius larger than 1 nm. Zhou et al.
17
have used the first principles cluster model
within the local density approximation to evaluate the mechanical properties of a
SWNT. The estimated values for tensile modulus, tensile strength, and Poissons
ratio were 0.764 TPa, 6.248 GPa, and 0.32,
17
respectively. The binding energy of the
nanotube was less than that of graphite due to the curvature effect. Lier et al.
18
calculated
the tensile modulus of zigzag and armchair SWNTs using the ab initio multiplicative
integral approach (MIA), which is based on the energy of elongation of nanotubes in
a simple tension (which is not constrained laterally or in any other way). They found
that the moduli of SWNTs or MWNTs were larger than that of a graphene sheet. The
MD simulations showed that the fracture behavior of zigzag nanotubes was more brittle
than the fracture behavior of armchair nanotubes.
19
The formation of a local Stone-
Wales defect (5775) in the deformed armchair nanotube induced ductile deforma-
tion.
FIGURE 14.6 Bond structures and corresponding energy terms of a graphene cell.
Torsion
N
o
n
-
b
o
n
d
e
d
B
o
n
d
s
tr
e
tc
h
in
g
Bond-angle
variation
Carbon atom
Inversion
E K dR C d
Total i
i
i j
j
j
( ) +
( )

1
2
1
2
2
2

14.2.3 FINITE ELEMENT MODELING
In the past few years, the demand for the development of faster methods to compute
the mechanical properties of nanostructures has been increasing. The classical
shell theory has been judged as too simple and less than accurate because it is
limited by some unrealistic boundary conditions. The finite element modeling
(FEM) method associated with the molecular dynamics (MD) or equivalent con-
tinuum (EC) model has been recently adopted to calculate the mechanical prop-
erties of nanotubes. Odegard et al.
20
have developed an equivalent continuum tube
model to determine the effective geometry and effective bending rigidity of a
graphene structure. Molecular mechanics considerations (see Equation 14.6 and
Equation 14.7) were first used to determine linking forces between individual
carbon atoms. This molecular force field was simulated by using a pin-joint truss
model; i.e., each truss member represents the force between two atoms, as shown
in Fig. 14.7. Therefore, the truss model allows the accurate simulation of the
mechanical behavior of nanotubes in terms of atom displacements. As the nanotube
was subjected only to a uniaxial load,
20
the bond stretching and bond-angle vari-
ation energies (see Equation 14.8) were considered. The strain energy of the whole
system was used in the FEM computation to estimate the effective thickness of the
nanotube layer. It was found that the effective thickness of the nanotubes (0.69 and
0.57 ) was significantly larger than the interlayer spacing of graphite, estimated
to be about a0.34 .
Li and Chou
2122
have worked out the contributions of van der Waals interactions
between individual carbon atoms within nanotubes using the FEM truss model. The
relationships of the structural mechanics parameters EA, EI, and GJ and the molec-
FIGURE 14.7 Truss model of carbon nanotube structure.
Carbon atoms
Truss members
Single-walled carbon nanotube
ular mechanics parameters K
U
, C
T
, and A
W
are shown in Equations 14.7a, b, and d
for each truss member:
(14.9ac)
In Fig. 14.8, the dependence of the van der Waals force versus the distance between
two carbon atoms is plotted. Nonlinear truss elements were used in simulations,
since the force between two carbon atoms was also nonlinear. A uniaxial load was
applied uniformly at the ends of the nanotubes;
2122
the effects due to the end caps
were neglected. It was found that the Youngs modulus of the nanotubes increased
as the nanotubes diameter was increased. The Youngs and shear moduli of MWNTs
were in the range of 1.05 r 0.05 and 0.4 r 0.05 TPa, respectively. For SWNTs, the
Youngs modulus was almost constant when the nanotubes diameters were larger
than 1.0 nm. The average Youngs modulus of the zigzag nanotubes was slightly
higher than the armchair type. Also the Youngs moduli of MWNTs were generally
higher (a7%) than those of SWNTs.
14.3 FABRICATION PROCESSES
OF NANOTUBE/POLYMER COMPOSITES
To achieve desirable properties of nanotube polymer-based composites, control of
the fabrication process in order to produce well-dispersed composites is an essential
issue. Several parameters such as the selection of dispersion solvents, sonication
time, stirring speed, and ambient temperature are the keys to disperse the nanotubes
FIGURE 14.8 Van der Waals force versus the distance between two carbon atoms.
0.05
0.03
0.01
0.01
0.03
0.05
0.07
0.09
4.4 5.4
6.4 7.4
8.4
Distance between two atoms (armstrong)
v
a
n

d
e
r

W
a
a
l
s

f
o
r
c
e

b
e
t
w
e
e
n

i
n
t
e
r
c
a
t
i
n
g

a
t
o
m
s
(
k
C
a
l
/
m
o
l
e
.
n
m
)
3.4 9.4
EA
L
K
EI
L
C
GJ
L
A

; and
uniformly in the resin. Another crucial factor that influences the strength of the com-
posites is the alignment of the nanotubes. Sonication is one of the most popular ways
to disperse nanotubes into polymer-based resin. Park et al.
23
first demonstrated the use
of in situ polymerization to disperse SWNT bundles in the polymer matrix. A dilute
SWNT solution, typically around 0.05 wt% of nanotubes in dimethylformamide (DMF)
subjected to sonication for 1
1
/
2
hours in an ultrasonic bath (40 kHz) was followed by
mixing with the hardener. The long sonication time may cause the nanotubes to entangle
and form bundles, which may result in losing the benefit of their high-strength properties
in composite materials. Mukhopadhyay et al.
24
have reported that if the sonication time
is greater than 4 h, it may destroy most of the graphene layers of the nanotubes and
possibly cause the formation of junctions in the nanotubes.
Lu et al.
25
have found that the use of different chemical solvents as dilute solutions
would greatly influence the integrity of composites. Since DMF has a high boiling
temperature compared with ethanol and acetone, the solvent remains inside the resin
and lowers the chemical reaction process between the resin and hardener. The boiling
points of acetone, ethanol, and DMF are 56qC, 78qC, and 130qC, respectively. Even-
tually the overall mechanical and chemical properties of the composite decrease. In
Table 14.1, the Vickers hardness values of different nanotube/epoxy composites
mixed by using different types of solvent are shown. Obviously the mechanical
properties of the composites are directly affected by the melting temperatures of the
solvent.
In Fig. 14.9, FTIR spectra of different types of composites are shown. Spectra
b, c, and d represent acetone, ethanol, and DMF solvents, respectively, which were
used to disperse the nanotubes in an epoxy matrix. Spectrum a is a pure epoxy. The
most prominent feature in the spectra b, c, and d is the appearance of a new absorption
band located at a1650 cm
1
. Considering epoxide and hydroxyl groups are the only
two reaction groups in the epoxy molecule, the a1650 cm
1
band can be assigned to
an amino group formed by the intermolecular nucleophilic substitution of hydroxyl
at the amide functionality, which can therefore be used to estimate the relative amount
of the product of the cure reaction. The band locations are different in the spectra b,
c, and d, with values of 1645 cm
1
, 1649 cm
1
, and 1664 cm
1
, respectively. This
indicates that there are functional and curing differences among epoxy resins with
different solvent treatments that may be related to the variation of the mechanical
TABLE 14.1
Vickers Hardness Values of Different Samples
Determined from a Load Force of 100 g and a Dwell
Time of 15 s
Composition Solvent Vickers Hardness No.
CNTs/epoxy Acetone 18.0 0.11
CNTs/epoxy Ethanol 14.4 0.08
CNTs/epoxy DMF 7.7 0.10
Pure epoxy 17.8 0.06
properties discussed previously. Blanchet et al.
26
directly dispersed nanotubes into
polyaniline (PANI) using a sonication method. The nanotubes were first sonicated
in xylene and that dispersion afterward was sonicated in the DNNSA-PANI solu-
tion. It was found that all the nanotubes were well dispersed into the PANI and
enhancement of the electrical conductivity was achieved. Tang et al.
27
have also
introduced the use of a melt processing technique for making nanotube/high-
density polyethylene (HDPE) composites. The nanotubes and HDPE were pre-
melted together to form pellets, followed by feeding them into a twin-screw
extruder to make composites (Fig. 14.10). In their work, one difficulty was con-
trolling the uniformity of the nanotubes in the HDPE. Once the pellets were melted,
some of the nanotubes were trapped inside the injection head of the extruder and
entangled, forming bundles. Although research related to nanotube/polymer com-
posites has been done for more than a half decade, the detailed study on how to
fabricate well-dispersed and -aligned nanotube/polymer composites is still a critical
issue.
14.4 INTERFACIAL BONDING PROPERTIES
OF NANOTUBE/POLYMER COMPOSITES
14.4.1 EXPERIMENTAL INVESTIGATION
It has been recognized that high mechanical strength of nanotube/polymer compos-
ites can be achieved when all nanotubes are aligned parallel to the load direction,
and an external stress should be effectively transferred to the nanotubes. Jin et al.
28
and
Wood et al.
29
have studied the control of the alignment of nanotubes in polymer
matrices for making high-strength advanced nanocomposite structures. However, the
FIGURE 14.9 The representative FTIR spectra measured from the samples a to d.
4000 3500 3000 2500 2000 1500 1000
a
b
c
d
Wavenumbers (cm
1
)
A
b
s
o
r
b
a
n
c
e
investigation on the mass production and control of the alignment of such nanotubes
toward the predetermined direction is still under investigation. The stress transfer
mechanisms of different types of nanotubes in nanotube/polymer composites have
not been comprehensively discussed elsewhere. Wagner et al.
30
and Qian et al.
31
have studied the load transfer properties of the nanotube/polymer composites by
conducting many experimental investigations. They found the interfacial shear
strength between the nanotube and polymer to be as high as 43.3 MPa when
measured in a fragmentation test. However, in several experimental investigations,
it was obvious that a poor adhesion property in the nanotube/polymer composites
was found because of the pulling out of the nanotubes.
32
In addition, a weak attraction
between different shells inside the multiwall nanotube also caused failures initiated
only at the outer shells of the nanotubes. The latest experimental study conducted
by Cooper et al.
33
demonstrated that the shear strength between a nanotube and
epoxy matrix varied with the diameter, the embedding length, and the number of
shells of the nanotubes. Two sections of an epoxy matrix were formed with a
nanotube bridging across the sections. The two sections were pulled apart using the
tip of a scanning probe microscope. The maximum shear strength ranging from 38
FIGURE 14.10 Process of preparing MWNT/HDPE nanocomposite films.
27
CNTs
HDPE
pellets
(a)
(d)
(g)
(b) (c)
(e) (f)
(h)
Heat
Dough
Film
Composite
pellets
MPa to 376 MPa was recorded, and a high shear strength was reported for SWNT
ropes. A pull-out of nanotubes from a solid polyethylene matrix was demonstrated
using an AFM,
34
and the average interfacial stress required to remove a SWNT from
the matrix was about 47 MPa. This experiment was closer to a realistic situation
since all the nanotubes were physically bonded to the matrix and a typical pull-out
test was actually conducted. However, as described in the previous section, the
mechanical properties of different kinds of nanotubes may influence their bonding
strength to the matrix. To better understand the interfacial bonding behavior of the
nanotube and matrix, an improved experimental setup to perform pull-out tests for
different types and geometries of nanotubes is needed.
14.4.2 THEORETICAL STUDY AND MOLECULAR
DYNAMICS SIMULATION
The molecular dynamics (MD) simulation is a powerful tool to predict the interfacial
bonding properties among different shells in nanotubes, as mentioned in Section
14.2.2, and also between the nanotubes and surrounding polymer molecules. Liao
and Li
35
simulated a pull-out action of nanotubes in polystyrene (PS) matrices using
the commercial software Hyerchem
. The adhesion strength between PS molecules

and a graphene sheet was studied using a molecular mechanics model employing
an empirical MM+ force field. A random coil of PS [(-CH
2
CHC
6
H
6
-)
n
] molecules
with n 2, 4, 10, 20, 40, and 80 was constructed and located near the surface layer
(Fig. 14.11). The nanotubes were then subjected to a pull-out force to measure the
interfacial shear strength. In their study, the shear stress between the nanotubes and
polymer estimated from the simulation was about 160 MPa. An experiment was
then subsequently carried out by one of the coauthors of Reference 35 and a strong
interfacial adhesion between nanotubes and matrix was observed from electron
microscopy.
36
Lordi and Yao
37
used a force field-based molecular mechanics calcu-
lation to determine binding energies and sliding frictional stresses between pristine
nanotubes and a range of polymer substrates. They found that all of the substrates
studied showed more sliding friction on a nanotube surface than friction between
two graphene sheets and much more than that between MWNTs. The key factor in
forming a strong bond between the polymer and nanotube lies in the polymers
morphology, specifically its ability to form large-diameter helices around individual
nanotubes. Frankland et al.
38
studied the effect of chemical cross-links on the
interfacial bonding strength between a SWNT and polymer matrix using MD sim-
ulations. Their models were composed of single-wall (10,10) armchair nanotubes
embedded into either a crystalline or amorphous matrix. The nonbond interactions
within the polyethylene matrix and between the matrix and nanotubes were modeled
with Lennard-Jones 6-12 potentials. They found that even a relatively low density
of cross-links can have a large influence on the properties of the nanotube-polymer
interface. Ren and Liao
39
also found that pull-out energies are affected by the thermal
condition of composites. However, in those simulated studies, due to the reduction
of complexity, chemical interactions between the nanotubes and matrix were gen-
erally ignored. Only nonbond interactions, and electrostatic and van der Waals forces
were assumed.
In the early stages of nanotube-related research, many diverse results on the
interfacial-bonding characteristic between nanotubes and polymer-based matrices
were found. Xu et al.
40
noted that high interfacial shear stress between a MWNT
and an epoxy matrix was observed from a fractured sample. Wagner
41
first used the
Kelly-Tyson model that has been widely used to study the matrix-fiber stress transfer
mechanism in micron-size fiber composites to study the interfacial shear strength
between the nanotube and polymer matrix. Since it was found that the binding force
between inner layers is very low, and sliding failure always occurs, only a single-
wall system was of interest in his work. In his study, it was assumed that an externally
applied stress to a nanotube/polymer composite is fully transferred to the nanotubes
via a nanotube-matrix interfacial shear mechanism at the molecular level; i.e., the
length of the nanotubes (l) is larger than the critical length (lc). A single nanotube
cylindrical model was used to study the stress transfer properties of the composite
shown in Fig. 14.12. To consider the force balance in the composite system, the
following equation is formed:
(14.10)
FIGURE 14.11 A molecular model of a double-walled carbon nanotube in a PS matrix.
35

NT i NT NT
o i
NT
o i
d dx d
d d d d
+ ( )

j
(
,
\
,
(

2 2 2 2
4 44
j
(
,
\
,
(
where W
NT
is the interfacial shear strength between the nanotube and matrix, V
NT
is
the tensile strength of a nanotube segment of length dx, and d
o
and d
i
are outer and
inner diameters of the nanotube, respectively.
After integrating Equation 14.10 and considering the critical length of a typical
short fiber system in composite structures, the interfacial shear strength can be written as
(14.11)
where l
c
/d
o
is the critical aspect ratio of the nanotube and d
i
/d
o
is the diameter ratio.
As an external applied stress of 50 GPa was used in the literature, the interfacial
strength was calculated for critical length values of 100, 200, and 500 nm. It was
concluded that the interfacial shear stress was affected by several factors: (1) the
critical length, and (2) the outer diameter of the nanotubes. Increasing the diameter
of the nanotube resulted in increasing the interfacial shear strength at the bond
interface.
Recently Lau
12
conducted an analytical study of the interfacial bonding proper-
ties of nanotube/polymer composites by using the well-developed local density
approximation model described in Section 14.2.2,
10
and classical elastic shell theory
and the conventional fiber pull-out model. In his study, several important parameters
such as the nanotube wall thickness, Youngs modulus, volume fraction, and chiral
vectors of the nanotubes were considered. It was found that decrease of the maximum
shear stress occurs by increasing the size of the nanotubes. Increasing the number
of the walls of the nanotubes will cause (1) a decrease of Youngs modulus of the
nanotubes, (2) an increase of the effective cross-sectional area, and (3) an increase
in the total contact surface area at the bond interface and the allowable pull-out
force of the nanotube/polymer system. In Fig. 14.13 a plot of the interfacial shear
stress of SWNTs with different chiralities is shown. The figure shows that the
maximum shear stress of a zigzag nanotube (5,0) is comparatively higher than those
of chiral (5,3) and armchair (5,5) nanotubes.
FIGURE 14.12 Stress transfer study using Kelly-Tyson model.

NT NT
c
o
i
o
l
d
d
d
j
(
,
\
,
(

j
(
,
\
,
(
j
(
,
\
,
(
1
2
1
1
2
2
In this chapter, a critical review of the strength of nanotubes, fabrication processes,
and interfacial bonding properties of nanotube/polymer composites is given. It was
found that there are still many uncertainties in measuring the mechanical properties
of nanotubes due to the difficulty of conducting nanoscale property tests. MD and
theoretical analyses are mainly based on certain assumptions that may not be appli-
cable to a real situation in nanotube/polymer composites. Dispersion properties and
control of alignment of nanotubes are also important issues that govern the global
properties of the composites. Presently no one has reported a practical large-scale
manufacturing method to disperse and align nanotubes. This is the main factor
limiting enhancement of the properties of nanotube/polymer composites. However,
a good interfacial bonding between the nanotubes and matrix is also required. In
the last section, we summarized some recent work on this particular issue. Two open
problems are (1) whether the chemical bonding between nanotubes and matrix exists
or not, and (2) whether nanotubes still maintain their extraordinary mechanical,
electrical, and thermal properties in the presence of chemical bonding between the
nanotube and matrix. Enhancement of the mechanical properties of advanced com-
posite materials will require much further investigation, and this is definitely a
challenging area for the composite community.
PROBLEMS
1. Which factors affect the mechanical, electrical, and thermal properties of
nanotube/polymer composites?
FIGURE 14.13 Plot of interfacial shear stress versus different types of nanotube.
0
50
100
150
200
250
0 40 80 120 160 200
z (nm)
S
h
e
a
r

s
t
r
e
s
s

(
M
P
a
)
Zigzag (5,0)
Chiral NT (5,3)
Armchair (5,5)
2. What are the differences between the mechanical properties of single-
walled and multi-walled carbon nanotubes? Do all inner layers of the
nanotubes contribute to the strength of the multi-walled nanotubes?
3. Fabrication processes of nanotube/polymer composites are crucial factors
that control the integrity of the composites. Which parameters have to be
controlled or studied in making high-strength composites?
4. Based on your own understanding, which type of nanotube (MWNT or
SWNT) is better to use to make nanotube/polymer composites?
ACKNOWLEDGMENTS
This work was supported by Hong Polytechnic University Grant G-T 861, Graduate
Enhancement Fund from the University of New Orleans.
REFERENCES
1. Iijima, S., Helical microtubules of graphite carbon, Nature, 354, 5658, 1991.
2. Snow, E.S., Campbell, P.M., and Novak, J.P., Single-wall carbon nanotube atomic
force microscope probes, Appl. Phys. Lett., 80, 11, 20022004, 2002.
3. Kiernan, G., Barron, V., Blond, D., Drury, A., Coleman, J., Murphy, R., Cadek, M.,
and Blau, W., Characterization of nanotube based artificial muscles materials, Pro-
ceedings of SPIE, 4876, 775781, 2003.
4. Gao, Y. and Bando, Y., Carbon nanothermometer containing gallium, Nature, 415,
599, 2002.
5. Liu, C., Yang, L., Tong, H.T., Cong, H.T., and Cheng, H.M., Volumetric hydrogen
storage in single-walled carbon nanotubes, Appl. Phys. Lett., 80, 13, 23892391, 2002.
6. Yu, M.F., Files, B.S., Arepalli, S., and Ruoff, R.S., Tensile loading of ropes of single
wall carbon nanotubes and their mechanical properties, Phys. Rev. Lett., 84, 24,
55525555, 2000.
7. Demczyk, B.G., Wang, Y.M., Cumings, J., Hetman, M., Han, W., Zettl, A., and
Ritchie, R.O., Direct mechanical measurement of the tensile strength and elastic
modulus of multiwalled carbon nanotubes, Mater. Sci. Eng. A, 334, 173178, 2002.
8. Cumings, J. and Zettl, A., Low-friction nanoscale linear bearing realized from multiwall
carbon nanotubes, Science, 289, 602604, 2000.
9. Qi, H.J., Teo, K.B.K., Lau, K.K.S., Boyce, M.C., Milne, W.I., Robertson, J., and
Gleason, K.K., Determination of mechanical properties of carbon nanotubes and
vertically aligned carbon nanotube forests using nano-indentation, J. Mech. Phys.
Solids, in press.
10. Tu, Z.C. and Ou-Yang, Z.C., Single-walled and multiwalled carbon nanotubes viewed
as elastic tubes with the effective Youngs moduli dependent on layer number, Phys.
Rev. B, 65, 233407, 2002.
11. Lau, K.T., Gu, C., Gao, G.H., Ling, H.Y., and Reid, S.R., Stretching process of single-
and multiwalled carbon nanotubes for nanocomposite applications, Carbon, 42,
423460, 2004.
12. Lau, K.T., Interfacial bonding characteristics of nanotube/polymer composites, Chem.
Phys. Lett., 370, 399405, 2003.
13. Lau, K.T. and Hui, D., The revolutionary creation of new advanced materials: carbon
nanotube composites, Comp. Pt. B Engineering, 33, 263277, 2002.
14. Lennard-Jones, J.E., The determination of molecular fields: from the variation of the
viscosity of a gas with temperature, Proc. R. Soc., A106, 441, 1924.
15. Chang, T.C. and Gao, H.J., Size-dependent elastic properties of a single-walled carbon
nanotube via a molecular mechanics model, J. Mech. Phys. Solid, 51, 10591074,
2003.
16. Lu, J.P., Elastic properties of single and multi-layered nanotubes, J. Phys. Chem.
Solids, 58, 11, 16491652, 1997.
17. Zhou, G., Duan, W.H., and Gu, B.L., First-principles study on morphology and
mechanical properties of single-walled carbon nanotubes and graphene, Chem. Phys.
Lett., 326, 181185, 2000.
18. Lier, G.V., Alsenoy, C.V., Doren, V.V., and Geerlings, P., Ab initio study of the elastic
properties of single-walled nanotubes, Phys. Rev. B, 58, 20, 1401314019, 1998.
19. Nardelli, M.B., Yakobson, B.I., and Bernholc, J., Brittle and ductile behaviour in
carbon nanotubes, Phys. Rev. Lett., 81, 21, 46564659, 1998.
20. Odegard, G.M., Gates, T.S., Nicholson, L.M., and Wise, K.E., Equivalent-continuum
modelling of nano-structured materials, Comp. Sci. Tech., 62, 18691880, 2002.
21. Li, C.Y. and Chou, T.W., A structural mechanics approach for the analysis of carbon
nanotubes, Inter. J. Solid Struct., 40, 24872499, 2003.
22. Li, C.Y. and Chou, T.W., Elastic moduli of multi-walled carbon nanotubes and the
effect of van der Waals forces, Comp. Sci. Tech., in press.
23. Park, C., Ounaies, Z., Watson, K.A., Crooks, R.E., Smith, J., Jr., Lowther, S.E.,
Connell, J.W., Siochi, E.J., Harrison, J.S., and Clair, T.L., Dispersion of single wall
carbon nanotubes by in situ polymerization under sonication, Chem. Phys. Lett., 364,
303308, 2002.
24. Mukhopadhyay, K., Dwivedi, C.D., and Mathur, G.N., Conversion of carbon nano-
tubes to carbon nanofibers by sonication, Carbon, 40, 13691383, 2002.
25. Lu, M., Lau, K.T., Ling, H.Y., Zhou, L.M., and Li, H.L., Effect of solvents selection
for carbon nanotubes dispersion on the mechanical properties of epoxy-based nano-
composites, Comp. Sci. Tech., in press.
26. Blanchet, G.B., Fincher, C.R., and Gao, F., Polyaniline nanotube composites: a high-
resolution printable conductor, Appl. Phys. Lett., 82, 8, 12901292, 2003.
27. Tang, W.Z., Santare, M.H., and Advani, S.G., Melt processing and mechanical prop-
erty characterization of multiwalled carbon nanotube/high density polyethylene
(MWNT/HDPE) composite films, Carbon, 41, 27792785, 2003.
28. Jin, L., Bower, C., and Zhou, O., Alignment of carbon nanotubes in a polymer matrix
by mechanical stretching, Appl. Phys. Lett., 73, 9, 11971199, 1998.
29. Wood, J.R., Zhao, Q., and Wagner, H.D., Orientation of carbon nanotubes in polymers
and its detection by Raman spectroscopy, Comp. Pt. A, 32, 391399, 2001.
30. Wagner, H.D., Lourie, O., Feldman, Y., and Tenne, R., Stress-induced fragmentation
of multiwall carbon nanotubes in a polymer matrix, Appl. Phys. Lett., 72, 2, 188190,
1998.
31. Qian, D., Dickey, E.C., Andrews, R., and Rantell, T., Load transfer and deformation
mechanisms in carbon nanotube-polystyrene composites, Appl. Phys. Lett., 76, 20,
28682870, 2000.
32. Lau, K.T. and Hui, D., Effectiveness of using carbon nanotubes as nano-reinforcement
for advanced composite structures, Carbon, 40, 15971617, 2002.
33. Cooper, C.A., Cohen, S.R., Barber, A.H., and Wagner, H.D., Detachment of nanotubes
from a polymer matrix, Appl. Phys. Lett., 81, 20, 38733875, 2002.
34. Barber, A.H., Cohen, S.R., and Wagner, H.D., Measurement of carbon nanotube-
polymer interfacial strength, Appl. Phys. Lett., 82, 23, 41404142, 2003.
35. Laio, K. and Li, S., Interfacial characteristics of a carbon nanotube-polystyrene
composite system, Appl. Phys. Lett., 79, 25, 42254227, 2001.
36. Wong, M., Paramsothy, M., Xu, X.J., Ren, Y., Li, S., and Liao, K., Physical interaction
at carbon nanotube-polymer interface, Polymer, 44, 77577764, 2003.
37. Lordi, V. and Yao, N., Molecular mechanics of binding in carbon-nanotube-polymer
composites, J. Mater. Res., 15, 12, 27702779, 2000.
38. Frankland, S.J.V., Caglar, A., Brenner, D.W., and Griebel, M., Molecular simulation
of the influence of chemical cross-links on the shear strength of carbon nanotube-
polymer interfaces, J. Phys. Chem. B., 106, 30463048, 2002.
39. Ren, Y., Fu, Y.Q., Li, F., Cheng, H.M., and Liao, K., Fatigue failure mechanisms of
single-walled carbon nanotube ropes embedded in epoxy, Appl. Phys. Lett., in press.
40. Xu, X.J., Thwe, M.M., Shearwood, C., and Liao, K., Mechanical properties and
interfacial characteristics of carbon-nanotube-reinforced epoxy thin film, Appl. Phys.
Lett., 81, 2833, 2002.
41. Wagner, H.D., Nanotubes-polymer adhesion: a mechanics approach, Chem. Phys.
Lett., 361, 57, 2003.
15
Nanoscale Intelligent
Materials and Structures
Yun-Yeo Heung, Inpil Kang, Sachin Jain,
Atul Miskin, Suhasini Narasimhadevara,
Goutham Kirkeria, Vishal Shinde,
Sri Laxmi Pammi, Saurabh Datta, Peng He,
Douglas Hurd, Mark J. Schulz,
Vesselin N. Shanov, Donglu Shi,
F. James Boerio, and Mannur J. Sundaresan
CONTENTS
15.1 Introduction
15.2 Review of Smart Materials
15.3 Nanotube Geometric Structures
15.3.1 Structures of Carbon Nanotubes
15.3.2 Structures of Noncarbon Nanotubes
15.3.3 Designations of Nanotubes and Nanostructured Materials
15.4 Mechanical and Physical Properties of Nanotubes
15.4.1 Elastic Properties
15.4.2 Electrical Conductivity
15.4.3 Magnetoresistance
15.4.4 Piezoresistance
15.4.5 Electrokinetics of Nanotubes
15.4.6 Piezoelectric Properties
15.4.7 Electrochemical Effects
15.4.8 Nanotube Power Generation
15.4.9 Nanotube Contact Phenomena
15.5 Review of Nanoscale Sensors and Actuators
15.5.1 Simulation of Nanotube Structures
15.5.2 Nanotube Strain Sensors
15.5.3 Actuators Based on Nanoscale Materials
15.5.3.1 Carbon Nanotube Electrochemical Actuators
15.5.3.2 Thermally Activated Actuators
15.5.3.3 Piezoelectric and Nanotweezer Actuators
15.5.3.4 Shape Memory Alloy and Platinum Nanoscale
Actuators
15.5.3.5 Biological Molecular Actuators
15.6 Manufacturing of Carbon Nanotube and Nanofiber Intelligent
Materials
15.6.1 Synthesis of Nanotubes
15.6.2 Functionalization of Nanotubes
15.6.3 Casting Nanotubes and Nanofibers in Structural Polymer
Electrolyte
15.6.4 Carbon Nanotube Composite Strain Sensors
15.6.5 Carbon Nanotube-Based Biosensors
15.6.6 Carbon Nanotube and Nanofiber Hybrid Actuators
15.7 Future Directions for Intelligent Materials
15.7.1 Carbon Structural Neural Systems
15.7.2 High-Temperature Nanoscale Materials
15.7.3 Power Harvesting Using Carbon Nanotubes
15.7.4 Intelligent Machines
15.7.5 Telescoping Carbon Nanotubes
15.8 Conclusions
Problems
Acknowledgments
References
15.1 INTRODUCTION
Since the time of the design of the lever and wheel, it has been a goal of mankind
to improve the performance of engineered systems. To achieve large advances in
the performance of current systems starting from the simplest machine elements
to automobiles, bridges, buildings, aircraft, helicopters, wind turbines, environmental
monitoring systems, national security systems, and surgical and medical devices
will probably require that these systems become intelligent. Intelligent systems and
machines will be autonomous, self-sensing, able to resist vibration, redistribute
loads, and change shape in response to their environmental conditions, and therefore
preserve their integrity, increase their lifetimes, optimize their performance, reduce
the cost of ownership, and provide continuous safety during use. The critical ingre-
dient needed to make structures and machines intelligent is intelligent materials.
The problem is that current intelligent materials have limited capabilities and new
classes of practical and enabling intelligent materials with vastly improved properties
are needed. This chapter therefore tackles the problem of designing new intelligent
materials and takes the approach of using nanoscale components and developing
processes to build intelligent materials from the bottom up.
The terms intelligent and smart are often used interchangeably, and they can
describe both materials and structures. Smart materials are defined as having sensing
or actuation properties such as a piezoelectric response to dynamic strain that causes
electrons to move. Smart structures use smart materials for sensing and actuation in
analog or digital closed-loop feedback systems. Intelligent usually designates a more
complex level of material behavior or control, possibly by using miniaturized electronics
built into the material itself or by using a more advanced feedback system such as a
digitally controlled structure programmed with a rudimentary ability to learn and reason.
Smart structures can also be passive such as a wing designed with bend-twisting
coupling to protect against flutter, or semiactive such as a magnetorheologic damper
that uses a feedback signal to control the voltage field and viscosity of a fluid. Smart
materials and structures that are fully active use feedback signals, processors, amplifiers,
and actuators to apply forces to control the vibration or shape of a structure.
Present smart/intelligent materials and structures have fundamental limitations in
their sensitivity, size, cost, ruggedness, and weight that have restricted their widespread
application. To address the need for new intelligent materials and systems, researchers
are investigating nanotechnology and also intersecting nanotechnology with biomi-
metics. Nanotechnology can provide the route to build perfect materials starting at the
atomic scale, as in nature, while biomimetics can provide the inspiration and archi-
tecture for material system design. Multifunctionality is a universal trait of biological
materials and systems and is an important attribute when developing intelligent mate-
rials and structures. Biological systems have been designed by nature from the smallest
components upward over a long period of time, and they have material capabilities
unmatched by man-made systems. The concept of biologically inspired nanotechnol-
ogy, or bionanotechnology, can be described as the process of mimicking the chemical
and evolutionary processes found in nature to synthesize unique almost defect-free
multifunctional material systems starting from the nanoscale up. Bionanotechnology
is thus a new frontier that can lead to new generations of smart materials and intelligent
systems that can sense and respond to their environment.
Although we are at an early stage of research, nanoscale intelligent materials
including carbon nanotubes, inorganic nanotubes, compound nanotubes, nanobelts,
and other nanoscale materials have produced great excitement in the research com-
munity. This is because of their remarkable and varied electrical and mechanical
properties including electrical conductance, high mechanical stiffness, light weight,
electron-spin resonance, electrochemical actuation, piezoelectricity, piezoresistance,
contact resistance, Coulomb drag power generation, thermal conductivity, lumines-
cence, and the possibilities for functionalizing these materials to change their intrinsic
properties. These attractive properties also have the potential to allow development
of self-contained nanoscale intelligent materials based on analog or digital control
of the material using built-in nanoscale electronics. The broad area of nanoengineer-
ing of intelligent materials is discussed in this chapter, starting with material syn-
thesis, to the processing of nanostructured hybrid materials and then fabrication of
devices such as strain sensors, biosensors, and wet and dry actuators. The focus is
on nanotube-based materials that have novel strain sensing and force actuation
properties and the associated material processing steps needed to build macroscale
materials from nanoscale components, including digital control of the material behav-
ior. Nanoscale smart materials are important because they often possess improved
or different properties from macroscale materials due to their near-perfect construc-
tion, large surface area per volume, and quantum effects.
112
While nanoscale materials, and in particular carbon nanotubes (CNTs), have
extraordinary properties, utilizing these properties at the nanoscale and at the mac-
roscale is presenting major challenges for scientists and engineers. Thus far, nanoscale
intelligent materials have mostly been based on carbon nanotubes. A single-wall
carbon nanotube (SWCNT) is a superelastic crystalline molecule that can have a
length-to-diameter ratio of 1000 or more, and an enormous interfacial area of
500 m
2
/gm or more. Custom designing of multifunctional materials by mixing nan-
otubes with bulk carrier materials is now producing a range of new properties such
as piezoresistivity and electrochemical transduction, as well as enhanced strength,
modulus, toughness, hardness, and thermal and electrical conductivity, while simul-
taneously decreasing thermal expansion and permeability. Controlling and using
these properties for smart materials is dependent on understanding the composition,
topography, and processing relationships that define nanomaterial behavior. Major
difficulties in forming nanocomposite materials occur because CNTs do not easily
transfer load to matrix materials because their small size can affect the molecular
structure of the host material, and because van der Waals bonding is weak in the
axial direction of the nanotubes and cannot prevent intertube longitudinal slipping.
Also the van der Waals bonding is strong in the transverse direction, which makes
it difficult to transversely separate and disperse the CNT. A consequence of this
behavior is that bundles of parallel nanotubes are difficult to pull apart transversely,
but they can axially slip and therefore can act as defects in polymer smart materials.
Mechanically and chemically altering the nanotubes to disperse and to bond with the
host material is helpful, but this can adversely affect the electrical and transduction
properties of the material. In addition, nanotube electrochemical actuators must allow
ion conductivity over the surface of the nanotube while simultaneously transferring
actuation strain to and from the nanotube. While many materials have the possibility
for electrochemical actuation, the effect is negligible in bulk materials because most
of their atoms are in the interior. Conversely, in SWCNTs all the atoms on are the
surface and the electrochemical actuation effect is large. When the nanotubes collect
in bundles or ropes, the inner nanotubes are shielded from the ion exchange and the
nanotubes also slip on one another. Therefore, the critical factors in the design of
nanotube intelligent materials, and especially electrochemical actuators, are that the
nanotubes must be separated and aligned for electrical conductivity and ion exchange,
and the strain must be transferred from the nanotube to the host material. Otherwise
macroscale nanotube intelligent materials and actuators will be weak and inefficient.
There is also the possibility of using nanotube actuators at the nanoscale, but there is
not much published research in this new area.
In the broadest sense, the goal of nanoscale materials research is to develop
multifunctional intelligent materials. These materials would possess high strain-to-
failure value and high stiffness, be lightweight, and have a high bandwidth for
actuation or sensing. They would serve as strong structural materials that simulta-
neously have sensing, actuation, and other functional capabilities. There is no other
smart material available today that is also a structural material. Therefore nanotube
hybrid actuator and sensor materials could become an enabling technology for the
improvement of all kinds of dynamic systems including helicopters, reconfigurable
aerodynamic surfaces, launch vehicles, ultra-high energy density wireless motors,
and nanodevices such as biomedical nanosensors, surgical robots, active catheters,
and biomedical implants. Nanotechnology is attractive and exciting because it may
produce the most efficient smart materials ever made. On the other hand, nanoscience
is presently difficult to implement in practical applications. Various techniques for
processing nanoscale materials are presented in the literature, but many of these
are difficult to extrapolate to the macroscale or to use for large volume production
due to the impracticality of the processing method or cost. Moreover some state-
ments in the literature seem to inflate the potential of nanotechnology, which may
lead to a backlash that will impede nanotechnology research. Stating that carbon
nanotubes are 100 times stronger at one-sixth the weight of steel is misleading,
because nanotubes are nanoscale orthotropic molecules and steel is an isotropic
bulk material, and the two are not directly interchangeable. It would be more
realistic to compare the properties of a SWCNT with those of a typical commercial
reinforcing fiber used today. In Table 15.1, the elastic properties of the Hexcel IM8
carbon fiber
12
are compared to the elastic properties of a SWCNT. The elastic
modulus of the SWCNT is 3.3 times greater than the elastic modulus of the carbon
fiber, and the specific modulus of the SWCNT is 4.4 times greater than the specific
modulus of the carbon fiber. The strength of the SWCNT is 8.9 times greater than
the strength of the carbon fiber, and the specific strength of the SWCNT is 12 times
greater than the specific strength of the carbon fiber. Also the elastic modulus of a
unidirectional SWCNT-epoxy composite is 3.2 times greater than the elastic modulus
of a unidirectional carbon fiber-epoxy composite, and the specific modulus of a
unidirectional SWCNT-epoxy composite is 4.2 times greater than the specific mod-
ulus of a unidirectional carbon fiber-epoxy composite. Since a typical carbon fiber is
a bulk material that is 3643 times larger in diameter than the SWCNT, there is a
significant problem of processing the nanocomposite material to bring its nanoscale
properties to the macroscale. Another seeming overstatement appeared in a New
York Times article: Todays Visions of the Science of Tomorrow, January 4, 2003.
In the article, it was predicted that nanotechnology will lead to aerospace vehicles
with 98% less structural mass. It is interesting to speculate how long it will take the
aerospace companies to meet this prediction.
On the positive side, large advances in nanoscale materials development have
occurred. Small percentages of CNT are being incorporated into polymers, carbon
fibers, and metals for reinforcement, and these hybrid materials have shown signif-
icant improvements in elastic modulus or toughness, as compared to the host polymer
material. Spinning of fibers from carbon nanotubes has produced strong fibers that
can be incorporated into polymers to produce super-strong materials when the cost
of the fiber decreases. Electrochemical actuation in carbon nanofiber composites has
also been demonstrated, and this material could provide actuator capability to large
structures. An important contribution of this chapter is to present useful approaches
to tackle the problems of manufacturing nanoscale intelligent materials. Also pre-
sented are some ideas a little on the wild side for building futuristic intelligent
structures, biodevices, and structural health monitoring systems. Readers should also
pay attention to nonstructural and smaller material volume applications for nanoscale
intelligent materials including biosensing, surgical manipulators, and smart
machines, which are also discussed in the chapter. One such idea for applications is
a fluid form of an intelligent material that may have properties that can be activated
by contact with another material. There are potentially many types of nanotube-based
fluids that take advantage of the small size and ability to functionalize nanotubes.
T
A
B
L
E

1
5
.
1
P
r
o
p
e
r
t
i
e
s

o
f

R
e
i
n
f
o
r
c
i
n
g

F
i
b
e
r
s

a
n
d

T
h
e
i
r

C
o
m
p
o
s
i
t
e
s

w
i
t
h

6
2
%

F
i
b
e
r

V
o
l
u
m
e

F
r
a
c
t
i
o
n
M
a
t
e
r
i
a
l
D
i
a
m
e
t
e
r
D

[
n
m
]
D
e
n
s
i
t
y
d

[
k
g
/
m
3
]
S
t
r
a
i
n

t
o

F
a
i
l
[
%
]
M
o
d
u
l
u
s

Y
[
G
P
a
]
S
p
e
c
i
f
i
c

M
o
d
u
l
u
s
[
Y
/
d
]
U
l
t
.

S
t
r
e
n
g
t
h

S
[
G
P
a
]
S
p
e
c
i
f
i
c

S
t
r
e
n
g
t
h
[
S
/
d
]
S
W
C
N
T

F
i
b
e
r
1
1
.
4
1
3
4
0
!
1
0
a
1
0
0
0
0
.
7
4
6
a
5
0
3
7
.
3

e
3
H
e
x
c
e
l

I
M
8

c
a
r
b
o
n

f
i
b
e
r
1
2
5
1
0
0
1
7
9
0
2
3
0
4
0
.
1
7
5
.
5
8
9
3
.
1
2

e
3
R
a
t
i
o

S
W
C
N
T
/
I
M
8

e
p
o
x
y
1
/
3
6
4
3
0
.
7
5
!
5
3
.
3
4
.
4
8
.
9
1
2
1
1
2
1
7
3
.
2
0
.
0
0
2
9
0
.
0
8
3
7
.
4

e
5
S
W
C
N
T
-
e
p
o
x
y
1
2
5
2
a
5
6
2

(
c
a
l
c
u
l
a
t
e
d
)
0
.
4
8
I
M
8

f
i
b
e
r
-
e
p
o
x
y
1
5
3
7
1
7
6

(
m
e
a
s
u
r
e
d
)
0
.
1
1
5
2
.
7
6
1
.
8

e
3
R
a
t
i
o

S
W
C
N
T
-
e
p
o
x
y
/
I
M
8

e
p
o
x
y
0
.
8
1
5
3
.
2
4
.
2
N
o
t
e
:

B
o
l
d
i
n
g

i
n
d
i
c
a
t
e
s

.
C
o
p
y
r
i
g
h
t

2
0
0
6

b
y

T
a
y
l
o
r

&

F
r
a
n
c
i
s
Figure 15.1 depicts a NanoSpray
TM
advertisement. An intelligent material is sprayed
onto another material and a chemical reaction takes place. Possible applications of
the spray are left to the readers imagination. Some nanosprays are undergoing testing
at University of Cincinnati. We have found that the nanotubes can be dispersed and
sprayed as a liquid, or dispersed, dried, and sprayed as a nearly invisible gas. Ultra-
thin uniform smart coatings can be quickly sprayed on materials using this approach.
The remainder of this chapter is organized as follows. Smart materials and the
geometric structures and properties of nanotubes are reviewed. Then a review of
new types of sensors and actuators built using nanoscale materials is given. Next
methods of processing nanoscale smart materials to manufacture macroscale mate-
rials are discussed. In nanotechnology research, we talk about the properties of
nanoscale materials and the properties of macroscale materials built using the nanos-
cale component materials, and these two sets of properties can be very different. With
this background, we cover the design of nanotube sensors and hybrid actuators and
show some exciting results. Then we propose how to build future intelligent machines
and devices using nanotubes. Finally conclusions based on research are noted.
15.2 REVIEW OF SMART MATERIALS
Smart materials are solid-state transducers that can have piezoelectric, pyroelectric,
electrostrictive, magnetostrictive, shape memory alloy/ceramic, piezoresistive, elec-
troactive, and other sensing and actuating properties and lack moving parts. Piezo-
electric materials are the most important smart materials today.
1329
In the mid-1980s
piezoceramic PZT was developed and has a high strain energy density for actuating
and sensing of structures. Active fiber composite (AFC) materials using PZT fibers
were developed in 1997
2123
by Hagood and Bent at MIT. In 2000, a variation of
FIGURE 15.1 NanoSpray advertisement.
NanoSpray
Product description
An aerosol dispersed solution of
carbon nanotubes or nanofibers specially
functionalized with chemical coatings.
this material called the macrofiber composite (MFC) was developed by NASA.
24
The AFC/MFC material utilizes parallel piezoceramic fibers or ribbons in an epoxy
matrix with an interdigitated electrode pattern on the top and bottom faces of the
composite layer, but not touching the fibers. At the University of Cincinnati (UC),
the construction of the AFC material was modified to develop piezoelectric artificial
nerves for sensing of structures.
25
While AFCs represent a large advance in the
development of high-impedance smart materials, AFCs are still relatively heavy and
only partially load bearing; there is a strength mismatch between the AFC and host
material, and their strain is a0.2%. These are significant limitations for developing
smart structures for advanced applications. There has been progress in developing
higher strain single crystal piezoelectrics, relaxor ferroelectrics, and alkaline-based
piezoelectric materials, but the processing methods for these materials are not yet
scalable to commercial quantities, and the brittle heavy nature of the materials
remains.
In 1991, multiwall carbon nanotubes (MWCNTs) were discovered by Sumio
Iijima,
30
and since then carbon nanotube synthesis has been intensely studied.
3134
See Part 1 of this book for current research in nanotube growth. In 1996 the C60
fullerene and SWCNT were synthesized by Richard Smalley et al.
32
Coupling
CNTs to electronic circuits was then studied as cited in References 3543. In 1999,
electrochemical actuation of SWCNTs was developed by Ray Baughman et al.
4468
;
in 20002001 CNT ropes and aligned membranes were developed;
6972
in 2001
composite materials strengthened with CNTs were studied by methods including
Raman scattering
7390
; and in 2001 the piezoelectric effects in carbon nanotubes
(CNTs), boron nitride nanotubes (BNTs)
9195
(also see Chapter 21 of this book),
and silicon nitride nanotubes were studied.
96100
Silicon oxide and vanadium oxide
nanotubes were also studied. In 2002 electrochemical actuation of a single carbon
nanotube was studied
49,54
and improvements in CNT actuation
58
were made. In
2003 vanadium oxide nanotubes (VNTs) were used for actuation.
98
Zinc oxide
nanobelts that have piezoelectric properties are discussed in Reference 100 and in
Chapter 4 of this book. Noncarbon nanotubes are discussed in References 101114,
and piezoelectric and other nanotubes are discussed in References 115121. Nanotube
electronic properties are given in References 122135. These research efforts have
verified that CNT-, VNT-, and BNT-based materials have extraordinarily high
strength, super-elasticity, and sensing capability, all at the same time. These properties
make nanotubes possibly the materials of choice for designing future smart materials
and strain actuators. However, BNTs are not readily available yet and they have
piezoelectric strain coefficients (d
ij
) that are perhaps 20 times lower than existing
piezoelectric ceramic materials, and it is not possible to separate the chiralities of
BNT. VNTs have lower actuation properties than CNTs, but their cost is projected
to be lower. Zinc oxide nanobelts may have good piezoelectric properties and strength,
but they are not commonly available or used in composites yet. Because the BNT,
VNT, and zinc oxide materials may have lower strain than electrochemical CNTs,
and the materials are not as well developed, their potential as a bulk smart materials
seems not yet as promising as CNTs, depending on the application. Hopefully
piezoelectric nanoscale materials will become more available soon.
Based on theory, CNTs have large electrochemical actuation properties. In prac-
tice, bringing these properties to the macroscale is difficult. Transferring charge to
the tubes in a SWCNT actuator and transferring the shear load through the actuator
have been limitations to realizing the theoretically predicted a2% strain and high
performance of SWCNT actuators and sensors. Ray Baughman and his group at the
University of Texas at Dallas and Matteo Pasquali and his group at Rice University
in Houston (Chapter 11) have described great progress in spinning CNT fibers. The
spun fibers are microns in diameter and the strength is greatly improved compared
to previous CNT fibers. A trade-off of strong fibers is that the nanoscale actuation
properties may be reduced compared to actuation of individual nanotubes because it
is difficult to intercalate ions into the interiors of the dense fibers.
To overcome the limitations of present smart materials and actuators, carbon
nanotube hybrid materials are being developed at UC with a focus on minimizing
ion blockage and shear lag, and increasing the electrochemical process velocity and
bandwidth of the material. Such CNT hybrid smart materials and the fiber-based
approaches mentioned are anticipated to enable new material systems with applica-
tions in mechatronic devices and large structures for defense and civilian systems.
Unlike other smart materials, these CNT hybrid materials and fibers can be unique
because they are simultaneously structural, functional, and smart. The CNT hybrid
material is predicted to have a large load-carrying capability, and high thermal and
electrical conductivity, actuation, and sensing properties. For design use, mathemat-
ical models of the CNT material must be developed in terms of the electrochemical
constitutive equations for a composite material using a solid polymer electrolyte
(SPE) matrix. In preliminary work at UC, a small CNT electrochemical actuator
material, a strain sensing material, and a power generation material have been built.
Moreover a multiwall carbon nanofiber (CNF) PMMA electrochemical actuation
material tested exhibited good actuation performance and is about 200 times lower
in cost than a SWCNT actuator material. Development of a SWCNT-MWCNT-CNF-
epoxy-SPE material that is a high-performance structural load-bearing material that
actuates and senses is the holy grail of the research described in this chapter. If
successful, it would provide actuation and sensing capabilities (i.e., intelligence) to
large structures.
15.3 NANOTUBE GEOMETRIC STRUCTURES
The geometries of different types of nanotubes are briefly described in this section.
The geometry plays an important role beyond the elastic properties the geometry
also defines the electronic properties of the nanotube and affects the sensing and
actuation properties of the smart material. Detailed nanotube mechanics are dis-
cussed in References 136142, and in Part 3 of this book.
15.3.1 STRUCTURES OF CARBON NANOTUBES
The crystal structure of graphite is layered and the carbon atoms within a layer
are strongly bound to their neighbors at room temperature, but the layers are
only loosely coupled by van der Waals forces, which makes graphite quite easily
deformable in a particular direction. This is why graphite is used as the cores of
pencils and for dry lubricants. Although nanotubes grow axially, a SWCNT can
be most easily described as a rolled-up tubular shell of graphite sheet with the
carbon atoms covalently bound to their neighbors. The bonding mechanism in a
carbon nanotube system is similar to that of graphite, which is sp
2
hybridization.
One of the properties in sp
2
hybridization is the occurrence of V-bonds and S-
bonds. The V-bonds are strong covalent bonds that bind the atoms in the plane
and result in the high stiffness and high strength of a CNT. On the other hand,
the S-bond is the interlayer interaction of atom pairs and is much weaker than
the V-bonds. One of the exciting properties of nanotubes relates to their electronic
band structure, which depends on the helicity and diameter of the nanotube. The
armchair nanotubes are metallic, whereas the zigzag and chiral tubes can be either
metallic or semiconducting. These results are major driving forces in the evolution
of nanoelectronics including piezoresistance of nanotubes,
143147
electrophoresis
and dielectrophoresis of nanotubes,
148156
and carbon nanotube flow sensors.
157158
Functionalization is the chemical alteration of the surface properties of nanotubes
for the purpose of improving adhesion to a matrix and improving the sensing
165177
and actuation properties of nanotubes.
4468
Apart from unique electronic proper-
ties, mechanical characteristics have provided excitement because they include
high strength, high stiffness, low density, superelasticity, and structural perfec-
tion. These remarkable multifunctional properties of CNTs provide the potential
to be used as smart materials for reinforcing, sensing, and actuating polymer
composites.
15.3.2 STRUCTURES OF NONCARBON NANOTUBES
Nanotube materials can be categorized as organic (e.g., carbon and various peptides)
and inorganic (e.g., silicon nitride, boron nitride) nanotubes. Inspired by the remark-
able functions of tubular structures in biology, much research on organic nanotubes
has focused on the biological functions in vivo and synthesis of various organic
nanotubes for in vitro environments.
178191
Special attention is needed to understand
noncovalent processes of tubular materials such as lipic, peptidic, or steroidic sys-
tems in organic nanotubes. Since organic nanotubes have remarkable functions in
biology, researchers have been trying to understand the role of organic nanotubes
and synthesize them in vitro. Organic nanotubes were studied for biosensor appli-
cations. In addition, remarkable functions of organic nanotubes (e.g., motor proteins)
can inspire researchers and may produce new types of biomimetic nanotube struc-
tures for sensors and actuators in the near future.
191
The small size, high strength, and remarkable physical properties of CNTs are
attractive, but nanotube properties depend on their diameter, thickness, helicity, and
defects, and the random helicity and high synthesis cost for purified samples are
barriers to applications. Therefore, researchers are increasingly investigating non-
CNT materials or inorganic nanotubes and nanotubes of different inorganic mate-
rials. Using our understanding of CNT phenomenona as a background, the extraor-
dinary properties of inorganic nanotubes can be predicted in such examples as
B
x
C
y
Z
z,
, and dichalcogenides of transition metals including M
e
S
2
(M
e
Mo,W, Nb,
S = S, Se, Te), MoO
3
, SiO
2
, Al
2
O
3
,V
2
O
5
, TiO
2
, NiCl
2
, InS, and Bi.
147154
Also, lead
titanate (PbTiO
3
, PT), lead zirconate (PbZrO
3
, PZ), and lead zirconium titanate
(PbZrO
3
-PbTiO
3
, PZT) solid solution nanotubes were synthesized by a chelate sol-
gel method.
110
These inorganic nanotubes and compound nanotubes can compensate
for the lack of control of chirality and properties of smart carbon nanotube struc-
tures, making it possible to modify the electrical, optical, mechanical, charging
pattern, and other properties of nanotubes over a wide range.
115121
Therefore we
cannot exempt the likelihood of using inorganic nanotubes for developing smart
materials. In particular, the piezoelectric properties of boron nitride nanotubes
(BNTs) were evaluated by calculating the polarization of a nanotube by the MNDO
(neglect of diatomic differential overlap method) which is a semiempirical method
for the calculation of molecular polarizabilities and hyperpolarizability. All types
of dichalcogenide nanotubes remain semiconducting even at the smallest diameters,
and BNTs are dielectric or semiconducting (n- or p-doped). Generally the width
of the band gap increases as the diameter of the NT increases, but in the case of
BNsT and MeS
2
, the band gap remains constant. Another interesting feature of
heteropolar non-CNT nanotubes is the charge transfer resulting in ionic interatomic
bonding. Compared to CNT, these non-CNT nanotubes are of great interest and
are enlarging the research field. More broad research not only in synthesis but also
in determination of inorganic nanotubes and their properties in bundles and in
polymers is needed to develop useful smart materials.
15.3.3 DESIGNATIONS OF NANOTUBES
AND NANOSTRUCTURED MATERIALS
Many types of nanotubes are being developed, and the terminology used to
describe them is not universal. Table 15.2 lists some of the different types of
nanotubes and and relevant references. The acronym for a carbon nanotube is CNT
and they are described as single wall (SW), double wall (DW), and multiwall
(MW). The nanotube may be amorphous or crystalline. Bamboo nanotubes have
internal closeouts along their length and this occurs in boron nitride nanotubes
(BNTs) and in some MWCNTs due to the growth mechanism. There are also
carbon nanofibers (CNFs), which are different from MWCNTs because in the
nanofiber the graphene walls are not parallel to the fiber axis. They are at an angle
of 20 in a Dixie cup arrangement and terminate in groups at the outer wall of
the fiber. This produces a rough surface and also exposes the ends of the tubes to
the surface, which may affect their electrical conduction and electrochemical
properties. SWCNTs have one wall and diameters of about 1.4 nm; MWCNTs
have multiple concentric parallel walls and diameters of 10 nm and larger; CNFs
have multiple concentric parallel walls at an angle to the fiber axis and have
diameters of 70 nm or larger. For comparison, graphite fibers used in traditional
composite materials have diameters on the order of 6000 nm.
Some of the important differences in the nanotubes/nanostructures are discussed
next. There are also other nanostructured materials including nanobelts, nanohorns,
nanoribbons, nanowires, nanoplates, nanonails, etc. that are not common and are
mostly under development. The type of nanotube affects the processing method for
smart materials; SWCNTs are the smallest in diameter, the most difficult to process
into smart materials, and the most expensive. MWCNTs do not have as high or
varied properties but are easier to process because of their larger diameter (a20 nm
dia.). CNFs are similar to MWCNTs but there is a critical difference: The nanofibers
are not continuous tubes and their surfaces show steps at the termination of the
walls of groups of fibers. The nanofibers include PR 24 HT (a70 nm diameter) and
the PR 19 (~130 nm diameter). The PR 19 has a CVD layer with a graphitic nature
and these fibers may be more robust to breakage, but the electrical properties of
the nanofiber are changed by the carbon coating. The PR 24 does not have a CVD
coating. There are low- and high-density variations of these two nanofiber types.
More information can be obtained from the manufacturer.
3
There are three main
types of carbon nanotube raw materials that have potential for use as electrochem-
ical actuators and smart materials: (1) SWCNTs, which have excellent electrochem-
ical properties. The cost, however, is very high, a$500/gm, and incorporating the
nanotubes into polymers at high loadings is difficult. (2) MWCNTs, which have
good electrochemical properties. The cost is high, a$150/gm bulk or $1000 for a
1 cm
2
array. Incorporating the nanotubes into polymers might be done by growing
arrays of nanotubes and casting the polymer around the arrays. (3) The PR 24 CNF.
The electrochemical properties are good, the cost is low, a$0.25/gm, and incorpo-
rating the nanofibers into polymers is easier because the fibers are large. All three
materials separately and in combination are being used at the UC to develop
intelligent materials. SWCNTs can be purchased commercially
12
or can be grown
using a commercial nanofurnace
5
; MWCNTs can be purchased commercially
2
and
also grown in arrays
5
; and CNFs can be purchased commercially.
3
A nanotube
TABLE 15.2
Designations of Nanotubes
Descriptions and Sources
SWCNT: 0.32 nm D, 2001000 nm L; grown by catalyzed CVD and other methods; cost of purified
form is a$500/gram
1,10,13
DWCNT: 24 nm D, 150 micron L; grown by CVD of methane over cobalt nanoparticles supported
on porous MgO nanoparticles
2,12
MWCNT: 1050 nm D, 150 micron L; grown by CVD; also bamboo MWCNT 2040 nm D, 120
micron L, internal closeouts
2,10
CNF (Pyrograf III
): 70200 nm D, 50100 micron L; nonconcentric nested tubes with walls angled

20 to the longitudinal axis, carbon crystal structure, produced as a vapor by decomposing
hydrocarbons with a catalyst, hydrogen sulfide, and ammonia, different grades; cost ~$100/lb
3
BNT: 340 nm D, 4080 nm D bamboo/cone nanotubes and nanosize polyhedral particles in clusters
15
SiCNT: 200 nm D, 60 micron L, amorphous nanocrystalline
11,141142
SiOx Nanostructure: 2030 nm D tubes, bundles, amorphous, thermal evaporation
2
Vanadium Oxide Nanotube (VNT): 15100 nm D, MW
144
synthesis laboratory using chemical vapor deposition (CVD) is a feasible way to
make nanotube materials for experimental purposes, (See Chapter 5). The differ-
ences in nanotubes affect the processing and performance of the smart material.
The dispersion and functionalization processes depend on the type of nanotube.
Dispersion of the nanotubes, preventing reagglomeration during curing, and the
effect the nanotubes have on the polymer structures may be the most critical part
of integrating any of the types of nanotubes into polymers.
15.4 MECHANICAL AND PHYSICAL PROPERTIES
OF NANOTUBES
Because of the great properties of nanotubes, smart materials developed using
nanotubes have the potential to improve the way we generate and measure motion
in devices from the nano- to the macroscale in size. The advantages of using
SWCNTs, MWCNTs, and CNFs are discussed here for building smart materials.
One important property that is ideal for all CNT materials is the ability to function-
alize (chemically modify the surface of ) the CNT and alter its properties. SWCNT
properties are discussed in the next section relative to forming CNT hybrid actuators
and sensors. In general, the elastic and transduction properties are the greatest for
SWCNTs. The properties decrease somewhat for MWCNTs and decrease further
for PR 24 CNF. Finally graphite fibers have almost no actuation properties, although
they can be used as electrical resistance-based sensors.
15.4.1 ELASTIC PROPERTIES
The Youngs modulus of SWCNT is as high as 1 terapascal. Compared with
aluminum, steel, or titanium, the strength-to-weight ratio of nanotubes in the
axial direction is much greater. Compared to a composite with carbon fibers, the
theoretical specific modulus of a SWCNT composite may be about four times
greater, as shown in Table 15.1. The maximum strain of a SWCNT is 10% or
more, which is higher than most high-strength structural materials. All of these
strong mechanical properties are due to the CC covalent bonding and the seam-
less hexagonal network. Thermal conductivity is also very high in the direction
of the nanotube axis. The mechanics of nanotubes is discussed in detail in Part
3 of this book. The incorporation of nanotubes into polymers can initiate multi-
functional improvements in the following areas with industrial applications noted
in parentheses: (1) increased physical strength (in aircraft and engine structural
components); (2) increased vibration damping (in aircraft structures, brake disks);
(3) resistance to surface wear and scratching (in brake disks, leading edge com-
ponents in aircraft, paint); (4) higher stiffness and dimensional stability (space
structures and structural components); (5) increased thermal conductivity (in
brakes, engines, spacecraft, electronics heat sinks, PWBs); (6) EMI shielding
(defense electronics); (7) energy harvesting from active nanocomposites (future
application for spacecraft, biomedical materials, communications); (8) increased
fracture resistance (brake disks, due to FOD in engines); (9) increased actuation
performance (future application for morphing of wings); (10) sensing for health
monitoring (future application with carbon-based structural neural system at the
UC); (11) corrosion and erosion resistance (aircraft edges); (12) clear nanocoat-
ings and materials (on anodized, aluminum, and polished surfaces, windows);
(13) increased electrical conductivity and photoluminescence (in ight vehicles);
and (14) actuation of structures (shape changing).
15.4.2 ELECTRICAL CONDUCTIVITY
Electronically, a carbon nanotube can be either metallic or semiconducting. Carbon
nanotubes also have been shown to conduct current ballistically without dissipating
heat. In theory, there are two propagating eigenmodes for a SWCNT. Mismatch of
the eigenmodes between the nanotube and a mechanical conductor makes electrical
contact difcult and high contact resistance may occur. The conductance (the
inverse of resistance) of SWCNT is predicted to be 2G
0
independent of the diameter
and length, where G
0
= 2e
2
/h = 1/12.9 k, which is one unit of the conductance
quantum, and e and h are the charge on one electron and Plancks constant,
respectively. This means that SWCNTs are predicted to have a minimum resistance
of about 6500 ohms, independent of their length. Temperature and magnetic elds
affect the resistance of the nanotubes. Metallic SWCNTs behave as long ballistic
quantum conductors with the charge carriers exhibiting a large phase-coherence
length. The semiconducting nanotubes are predicted to have the strongest electro-
chemical properties, but this is difcult to verify in macroscale actuators. The
properties of the nanotubes in a structural polymer electrolytes (SPEs) dene the
electrical and ion exchange and actuation performance. The metallic or semicon-
ducting properties of SWCNTs actually depend on their helicity or chirality, which
is denoted by a pair of integers (n, m). In particular, they are predicted to be metallic
if n m = 3q, where q is an integer. While armchair CNTs are metallic, electrical
properties of chiral and zigzag CNTs vary with the diameter. Therefore rational
design of nanotube smart structures requires a fundamental understanding of the
conductive properties of nanotube materials and how conductivity depends on tem-
perature and dimensionality.
122123
Based on recent observations of electron transport
and both theoretical and experimental results, SWCNTs are ballistic in nature,
implying the absence of inelastic scattering.
125
Experimental results from metallic
SWCNTs exhibited resistance below 100 k at room temperature, approaching the
theoretical lower limit of 6.5 k.
126
Transport in MWCNTs is controversial regarding ballistic and diffusive con-
ductivity.
172
In experiments, Bachtold et al. showed that resistance per unit length
of MWCNTs was above 5 k/m, resulting in diffusive conductivity.
122
Frank
et al. demonstrated that conductivity of MWCNTs was ballistic and transport was
conned to the outer layer of the tube.
128
C. Gerger et al. reported that resistance
of MWCNTs was at most 200 /m. This implies that MWCNT conductivity is
ballistic with almost a 30-m mean free path. The most interesting result is that
only the surface of MWCNT transport current, and high current densities can be
applied.
129130
Since our interest is in macroscopic smart nanotube structures, the
conductivity properties of SWCNTs and MWCNTs must be brought to the micro-
scale or macroscale in the form of composites. Then the transport features of CNT
composites can be considered for optimal design of nanotube smart materials.
132133
In ropes or bundles of SWCNT or in mats consisting of randomly oriented ropes,
the resistance increasing with temperature has been observed. M. Radosavljevic
et al. found that the coupling between nanotubes in the bundle is weak, and the
current is carried predominantly by nanotubes on the surface of the bundle at low
fields.
134
At high bias voltage just before complete breakdown, they observed an
additional current increase caused by the coupling to the nanotubes in the interior
of the bundle. These results indicate that nanotubes that are bundled may have
different transduction properties compared to individual nanotubes.
15.4.3 MAGNETORESISTANCE
The CNTs also have spin-dependent transport properties or magnetoresistance. The
direction of magnetization of the ferromagnetic electrodes used to contact the nano-
tube defines the spin direction of the charge carriers into and out of the nanotube
and a change in the resistivity of the nanotube. Spintronic nanoscale devices in theory
can be built using the superconductivity and magnetoresistance effects, where the
nanotubemetal junction appears to have a strong effect on the spin-dependent
transport. When a magnetic field is applied perpendicularly to the nanotube at low
temperature, the resistance shows a change of 15%. Also a strongly suppressed con-
ductance gap at zero bias voltage on a substrate can open at low temperature (4.2 K).
The magnetoresistance effect would be difficult to measure for sensing/actuation
of strain of the nanotube and for use in a smart material.
15.4.4 PIEZORESISTANCE
Piezoresistance is the change in electrical resistance with strain and is a useful
property for self-sensing of intelligent structures. Piezoresistance is discussed in
References 143147. A pioneering experiment showed that the conductance of a
metallic CNT could decrease by orders of magnitude when strained by an atomic
force microscope tip.
147
It was shown that the band structure of a CNT can be
dramatically altered by mechanical strain and the conductance of the CNT can
increase or decrease based on the chirality of the nanotube. This happens because
the strain changes the structure of the quantum states available to the electrons.
Metals conduct electricity easily because their electrons have easy access to the
quantum states that carry the electrons long distances. These states are in the
conduction band of the electronic structure. In semiconducting nanotubes, there is
a band gap that serves as an energy barrier that electrons must overcome to reach
the conduction band. The extra energy push to overcome the band gap can come
from heat or an electric field or strain. Actually strain changes the band structure,
which changes the electrical properties, making the nanotube or nanocomposite
material more or less conductive (piezoresistive), depending on the chirality of the
nanotube, the polymer, and the percent loading of the nanotubes.
80,82
Raman spectroscopy is a very useful tool to characterize carbon materials
because it gives information about the size of graphite crystallites and the degree of
ordering of the material.
166
Of particular importance, Raman shifts can represent all
types, shapes, and orientations of nanotubes. Therefore, Raman spectroscopy was
used to study the vibrational modes of carbon nanotubes and nanoropes and to
determine the characteristic properties of armchair, zigzag, and chiral tubes. It is
also reported that a polymer embedded with low-weight fractions of carbon nano-
tubes becomes Raman-active under an applied mechanical strain. Small amounts of
SWCNTs embedded in a polymer matrix were used to sense the mechanical response
of the polymer using Raman spectral data. The polarized Raman methods were very
useful tools for mapping the stress in a polymer plate under uniaxial tension. This
could be one way to sense in a smart structure, but there is still a limitation since a
large and expensive Raman spectroscope instrument is required.
The electronic and mechanical degrees of freedom are tightly coupled in the
CNT system. Many researchers tried to prove how bond stretching and twisting in
nanotubes affects the electrical properties of nanotubes, since the electromechanical
properties of SWNTs depend on nanotube chirality (m, n). These properties have
been exploited to sense mechanical deformations by a change in conductance or
resistance of the nanotube due to the potential application for nanoscale smart
structures such as NEMS devices. Several groups calculated conductance variations
due to mechanical deformation theoretically. Even for small bending angles, the
conductance changed, and the conductance of metallic SWCNTs was lowered up to
tenfold at the a45q bending angle.
147
Experimental investigation of nanotubes was
carried out with suspended nanotubes that were manipulated with an AFM tip while
the resistance or conductance was monitored.
147
Conductance of the SWCNT
decreased each time the AFM tip pushed the SWCNT down but recovered as the
tip retracted. These observed changes are entirely due to the mechanical elastic
deformation of the SWCNT. This phenomenon can be explained by the changes in
local bonding configuration from sp
2
to nearly sp
3
. As the AFM tip pushes the
SWCNT down, local S-electron density is decreased significantly. Drastic reduction
in S-electron density causes a significant decrease in conductance, because local S-
electron density, which is delocalized, is responsible for the electrical conduction.
Reversibility in the process and the dramatic variation in conductance are good
properties of reversible electromechanical transducers for smart structures. In par-
ticular, Z.L. Wang et al.
120
reported that the conductance is quantized and it is
independent of the length and width of the carbon nanotube. This property could
be used to build smart sensors for health monitoring systems. Jien Cao et al.
171
showed that small band-gap semiconducting nanotubes exhibited the largest resis-
tance changes and have good piezoresistive gauge factors under axial strains. This
idea can be used for nanotube-based smart structures.
15.4.5 ELECTROKINETICS OF NANOTUBES
In fluids, ponderomotive responses of particles can be produced by externally applied
time-dependent electrical fields. The electrical properties (conductivity and dielectric
constant) of a nanotube are usually different from those of a fluid. Therefore when
a nanotube is in an electrolyte, it will attract ions of opposite electrical polarity,
forming an electrical double layer. If a uniform DC electric field is applied to
nanotubes suspended in an electrolyte, the electrical double layer surrounding the
nanotube is distorted, and electrical charges that define the nanotubes structure are
induced to appear at the interfaces.
148156
Lateral movement may occur if the nano-
tube has a net electrical charge. This movement is called electrophoresis, which is
the attraction and movement of charged particles due to an electric field in a fluid.
If an AC field is used, the electrophoresis effect is predicted to become small at
frequencies above about 1 KHz.
The distortion of the electrical double layer and the creation of interfacial charges
cause the electric dipole moment and this allows the nanotube to be moved in an
electric field. These distortions of the electrical double layer have a finite relaxation
time and this effect becomes small at a much higher frequency exceeding about 50
KHz. If the nanotube is polar, the dipole moments due to polarization have an effect
up to MHz frequencies. The total lateral electric force acting on a nanotube of net
charge Q in a nonuniform electric field vector E is F QE (m) E where is
the Del operator that defines the field gradient and m is the dipole moment vector.
If the electric charge is zero or the frequency exceeds about 1 KHz, the Coulombic
interaction effect is small and the dipole moment and field gradient will dominate
the motion of the nanotube. The time-averaged force acting on the nanotube is given
by the real part of the dipole moment and is called dielectrophoresis, which is
the attraction and movement of uncharged particles due to polarization induced by
nonuniform electric fields in a fluid. The magnitude of the dipole moment caused by
induced electrical polarization depends on the size of the nanotube and the complex
conductivities of the nanotube and the suspending fluid medium. The dielectro-
phoretic force itself depends on the size and shape of the nanotube. The magnitude,
polarity, and time response of the dipole moment induced in a nanotube in an imposed
electric field provide a means to understand how electric fields can be used to
manipulate nanotubes using electrophoresis, dielectrophoresis, electrorotation, and
traveling fields. Electrorotation is the rotation of a nonuniform electric field to induce
a torque on a particle causing the particle to rotate. Traveling wave dielectrophoresis
is the introduction of a traveling electric field of specified wavelength for the selective
manipulation, trapping, and separation of particles using a frequency variation of
the applied dipole moments. Based on these properties, dielectrophoresis may have
use in separation of zigzag nanotubes from semiconducting CNTs to build actuators
and sensors and for aligning or dispersing CNT in solvents or resin systems. How-
ever, the forces involved in the movement of CNTs using electrophoresis and dielec-
trophoresis are very small and these effects are not considered suitable for high-
performance actuation applications as in smart structures.
15.4.6 PIEZOELECTRIC PROPERTIES
Piezo-, pyro-, and ferroelectric materials have been in use for two decades, and many
applications of piezoelectrics for sensors and actuators in smart structures have been
developed. Some of the highlights include electrostrictive materials for nonlinear
actuators, ferroelectric single crystals with very high electromechanical couplings
for medical transducers, thin and thick PZT films for MEMS, and multilayer-type
actuators and sensors for smart structures and structural health monitoring.
1329
Piezoelectric materials have electromechanical coupling and are used for the health
monitoring of structures since they can be embedded or attached to flexible structures
as sensors or actuators. Recently it has been predicted that CNTs and BNTs exert
piezoelectric effect, and there is some possibility for using these to develop smart
structures. (See Chapter 21.)
N.G. Lebedev et al.
115121
modeled the piezoelectric effect in SWCNTs and BNTs
based on quantum chemical calculations of nanotube polarization in accordance with
the stretch of nanotubes along their axis. They showed that the piezoelectric values
are very small for CNTs and increase in accordance with the increasing diameter
for BNTs. The physics group of North Carolina State University recently presented
polarization and piezoelectricity studies of BNTs. They reported the existence of
nonzero spontaneous polarization fields since most BNT are noncentrosymmetric
and polar. However contrary to their predictions, the combined Berry phase and
Wanier function analysis demonstrated that electronic and ionic spontaneous polar-
ization cancel exactly in systems. However after the intrinsic helical symmetry is
broken, their results suggested that the piezoelectric constants of zigzag BNTs are
larger than those in the PVDF polymer family. Piezoelectric properties of zigzag
BNT bundles and piezoelectric materials are also listed in Chapter 21.
Oxide-based pyroelectric and ferroelectric materials have the general formula
ABO
3
and are employed for their dielectric, piezoelectric, electrorestrictive, pyro-
electric, and electro-optic properties and their functional perovskite structure. Chang
et al. reported the synthesis and characterization of perovskite nanotubes, which are
lead titanate (PbTiO
3
[PT]), lead zirconate (PbZrO
3
[PZ]), and zirconium tianate
(PbZrO
3
PbTiO
3
[PZT]).
110
These nanotubes have the potential to be used as intel-
ligent materials for smart structures, active catheters, strain sensors, force actuators,
accelerometers, and other applications. Certain nanotubes such as boron nitride
nanotubes, as discussed, and zinc oxide nanowires and nanobelts are naturally polar
in bundles, and the phoretic effects and piezoelectric effects will be present. However,
the piezoelectric effect of BNTs is smaller than for piezoelectric ceramic materials,
and the piezoelectric effect of ZnO is not well known. Also the BNTs and ZnO are
not readily available. For the CNTs, the van der Waals, piezoresistive, and electro-
chemical effects are present while the piezoelectric effect is very small. Therefore
using piezoelectric nanotubes, wires, and ribbons is promising but currently is in an
early stage of developing high-strain actuator materials, as compared to using elec-
trochemical nanotubes.
15.4.7 ELECTROCHEMICAL EFFECTS
Introducing excess charge into CNTs produces mechanical deformations that do
mechanical work. The charge injected into the valence or conduction band causes
the electronic structure to shift. The electrochemical effect should produce up to 2%
strain
4468
based on the basal plane intercalation strain of graphite. The electrochem-
ical property can generate large strains and forces using low voltages. Therefore,
the electrochemical and piezoresistive properties of CNTs are considered promising
for actuation and sensing. The CNT electrochemical actuation mechanism is a
volume change of the nanotube induced by ion exchange when the nanotube is
immersed in an electrolyte. If a voltage source is connected to the CNT, anions and
cations in an electrolyte attach to the surface of a cathode CNT or anode CNT,
respectively, because of the electrical charging effect. Consequently the interface
surface of the CNT is expanded to produce an elongation of the actuator. This effect
has been under study using SWCNTs by Baughman et al. since 1999
45
and recently
at the UC.
6668
Research is also being done at UC to investigate the use of different
electrode geometries and applied voltage schemes for the remote actuation of
SWCNTs in electrolytes. Electrochemical actuation using CNFs is also being
studied at UC. The CNFs cannot produce macroscopic sheets because the van der
Waals forces are smaller than for SWCNTs. Even in the SWCNTs, the van der
Waals forces do not provide efficient shear transfer. Therefore, research is focusing
on developing a polymer host for nanotubes that can provide ion exchange for the
electrochemical effect.
Little research has been done using MWCNTs for electrochemical actuation.
A forest-like aligned MWCNT actuator in an aqueous electrolyte has been
tested,
46
but the actuation mechanism is an electrostatic repulsion among nano-
tubes. This electrical charging repulsion actuation mechanism produces a low
actuation force. A comparison of the electrochemical actuation effect of CNTs
with other actuator types is given in Table 15.3. The theoretical predicted energy
density of CNT actuators is two orders of magnitude greater than for piezoceramic
materials. The analogy of the piezoresistance effect in an electrolyte is called
electrochemical impedance, and this effect is being studied related to self-sensing
nanotube actuators.
66
15.4.8 NANOTUBE POWER GENERATION
Power generation is discussed in the literature.
157158
The flow of an ionic fluid over
SWCNT paper in the direction of the flow generates a charge. The voltage produced
fits a logarithmic velocity dependence, and the magnitude of the voltage also depends
on the ionic conductivity and the polar nature of the liquid. The dominant mechanism
thought to be responsible for this highly nonlinear response involves a direct forcing
of the free charge carriers in the nanotubes by the fluctuating Coulombic field of the
liquid flowing past the nanotubes. This sensor can be scaled down to length dimen-
sions on the order of micrometers, and the sensor has high sensitivity at low velocities
and a response time better than 1 ms. Therefore nanotube paper could also be used
to generate charge in a flowing liquid environment, which may be useful in biomedical
applications.
15.4.9 NANOTUBE CONTACT PHENOMENA
Contact phenomena between nanotubes or a nanotube and an electrode must be
considered, including conductivity phenomena of the nanotube when an optimal con-
figuration to design smart nanotube structures is needed. Compared to the contact
resistance of CNTs, the intrinsic resistance of CNTs is small. Therefore the contact
resistance was not significantly affected by the CNTs own properties such as defects or
diameter. Alper Buldum reported that conductance values with simple end-end contacts
are high with a negative differential resistance.
35
Experiments show that end contact
T
A
B
L
E

1
5
.
3
A
p
p
r
o
x
i
m
a
t
e

P
r
o
p
e
r
t
i
e
s

o
f

P
i
e
z
o
e
l
e
c
t
r
i
c

C
e
r
a
m
i
c

a
n
d

N
a
n
o
t
u
b
e

M
a
t
e
r
i
a
l
s
M
a
t
e
r
i
a
l
T
e
m
p
e
r
a
t
u
r
e
L
i
m
i
t

[
q
C
]
E
l
a
s
t
i
c
M
o
d
u
l
u
s
Y

[
G
P
a
]
B
a
n
d
-
w
i
d
t
h
[
H
z
]
V
o
l
t

R
a
n
g
e
D
e
n
s
i
t
y
[
K
g
/
m
3
]
P
k

S
t
r
a
i
n
H
m
S
t
r
a
i
n
/
V
o
l
t
H
m
/
v
o
l
t
s
S
t
r
a
i
n

E
n
e
r
g
y
[
J
/
m
3
]
E
n
e
r
g
y

D
e
n
s
i
t
y
[
J
/
K
g
]
P
i
e
z
o
-
I
n
d
u
c
e
d
S
t
r
e
s
s

e
3
3
[
C
/
m
2
]
P
Z
T

m
o
n
o
l
i
t
h
i
c

w
a
f
e
r
1
1
7
5
5
5
6
9
1
0
6
2
0
0
7
7
0
0
.
0
0
0
2
6
1
.
3

u

1
0
6
2
3
3
2
0
.
3
5
1
0
A
c
t
i
v
e

f
i
b
e
r

c
o
m
p
o
s
i
t
e
2
a
1
1
0
3
5
1
0
5
2
.
8
K
v
4
2
5
0
.
0
0
1
8
6
.
4

u

1
0
6
5
6
,
7
0
0
1
3
.
3
4
6
.
2
P
i
e
z
o
e
l
e
c
t
r
i
c

(
5
,
0
)

B
N
T
3
a
1
0
0
0
1
2
2
0
1
3
7
0
0
.
4

m
a
x
P
i
e
z
o
e
l
e
c
t
r
i
c

(
5
,
5
)

C
N
T
4
6
0
0
1
0
0
0
1
3
3
0
.
0
0
0
3

m
a
x
E
l
e
c
t
r
o
c
h
e
m
i
c
a
l

C
N
T
5

5
.
5
1
0
2
2
1
3
3
0
.
0
0
2
1

u

1
0
3
1
1
,
0
0
0
8
.
2
5
E
l
e
c
t
r
o
c
h
e
m
i
c
a
l

C
N
T
6
6
4
0
1
0
3
1
1
3
3
0
.
0
1
1

u

1
0
2
2
9

u

1
0
6
2
4
0
0
A
l
u
m
i
n
u
m

n
i
t
r
i
d
e

(
A
L
N
)
7
a
1
0
0
0
0
.
5
8
S
i
C
N
T

n
a
n
o
t
u
b
e
8
Z
i
n
c

o
x
i
d
e

n
a
n
o
b
e
l
t
9
a
3
9
6
5
V
2
O
5

n
a
n
o
t
u
b
e
1
0
1

P
Z
T

5
A
.
2

T
e
s
t
i
n
g
.
3

N
a
r
d
e
l
l
i
,

M
.

(
t
h
e
o
r
y
)
.
4

L
e
b
e
d
e
v
;

N
.

(
t
h
e
o
r
y
)
.
5

Z
h
a
n
g
,

Q
.
M
.

a
n
d

B
a
u
g
h
m
a
n
,

R
.

(
b
u
n
d
l
e
d

t
e
s
t
)
.
6

N
o
n
b
u
n
d
l
e
d

t
h
e
o
r
y
.
7

T
e
s
t
i
n
g
.
8

P
r
o
p
e
r
t
i
e
s

t
o

b
e

d
e
t
e
r
m
i
n
e
d
.
9

W
a
n
g
,

Z
.
L
.
1
0

B
a
u
g
h
m
a
n
,

R
.

a
n
d

S
p
a
h
r
,

M
.
Y
m
2
2
/
Y
/
m
2
2
C
o
p
y
r
i
g
h
t

2
0
0
6

b
y

T
a
y
l
o
r

&

F
r
a
n
c
i
s
resistance of armchair tubes is low due to a strong coupling between metal and
carbon atoms compared to weak coupling of the side contact. Reference 35 cites a
two-terminal nanotube junction formed by bringing two tubes ends together in
parallel. (l is the contact length.) The transmission coefficient T of the two-armchair
tube [(10-10)(10-10)] junction for l 564 was reported. The current-voltage
characteristics of the [(10-10)(10-10)] junction for l 546 and a model of a four-
terminal junction formed by crossing two nanotubes were also reported along with
contact resistance of the [(18,0)(10,10)] junction as a function of rotation angle T.
It was determined that contact resistance changes with the diameter of the nanotube,
the chirality of the nanotube, the Fermi wave vector of the metal, and the area of
contact. The contact resistance between nanotubes and various metals is important
and low ohmic contacts are needed to develop smart material actuator and sensor
systems. Based on the various experimental and theoretical results, metals with few
d vacancies in atomic structure such as Fe, Ni, and Co reveal more finite solubility
for carbon than Al, Au, and Pd, which have no d vacancies. This phenomenon is
due to curvature-induced rehybridization of carbon sp
2
orbitals with the Ni d orbital.
Also Ti and Nb, which are 3d and 4d metals with many d vacancies, form strong
bonds with CNTs.
4042
Although the Fe, Ni, and Co are better conductors, they are
corroded in an electrolyte. Palladium would not be corroded in an electrolyte. Silver-
filled conductive epoxy was shown to lower the contact resistance when contacting
copper with a carbon nanotube epoxy material in a dry environment, and this process
is used for making piezoresistive sensors at UC.
15.5 REVIEW OF NANOSCALE SENSORS
AND ACTUATORS
This section reviews the literature discussing the modeling and design of nanoscale
sensors and actuators.
15.5.1 SIMULATION OF NANOTUBE STRUCTURES
In order to simulate smart nanostructured materials, the electromechanical coupling
effect must be determined for the constitutive modeling of intelligent materials.
The ab initio method, tight-binding method, and classical molecular dynamics
(MD) are major categories of nanotube molecular simulation. The ab initio method
could provide the exact solutions of the Schordinger equation, but assumptions
and approximations are needed, and advanced algorithms and the duration time
for calculation are weak points for realistic modeling of nanotube smart materials.
Also only limited solutions are available for a specific class of problem. But the
ab initio method can give the most accurate results among all methods. Many
researchers have studied the mechanical, electrical, and optical properties of nano-
tubes based on the ab initio method.
136142
Nanotube mechanics is presented in Part
3 of this book and provides more detail. D. Snchez-Portal et al. used the ab initio
method for the study of structural, elastic, and vibrational properties of
SWCNTs with different radii and chiralities. M. Verissimo-Alves et al. obtained
electromechanical behavior based on ab initio calculations of charged graphene
and single-wall carbon nanotubes. Properties of non-CNT materials also were
calculated.
The tight binding method initially attributed to Slater and Koster in 1954 is a
semiempirical method for electronic structure calculations. The key role is that the
Hamiltonian is parameterized and simplified before the calculation, rather than
constructed it from first principles, and the total energy and electronic eigenvalues
are deduced from the Hamiltonian matrix. This is a very convenient method for
molecular simulation because it can consider larger systems than the ab initio
method. S. Reich et al. derived an analytic expression for the tight-binding dispersion
including up to the third-nearest neighbors, while V. Meunier presented the theoret-
ical modeling of scanning tunneling microscopy of nanotubes for computation of
STM images and current-voltage characteristics. The structures and electronic states
of capped carbon nanotubes were investigated using a tight-binding method by
Yusuke Kasahara. Also, A.A. Maarouf et al. presented a tight-binding theory to
analyze the motion of electrons between carbon nanotubes bundled into a carbon
nanotube rope.
142
In the case of non-CNTs, properties of charged borocarbide (BC)
nanotubes were calculated by the density functional tight binding method. MD is a
particle motion analysis based on Newtons second law. This method can consider
larger systems than previous methods. Nan Yao presented a detailed investigation
of the spring behavior of carbon nanotube caps based on MD simulation, and Jie
Han et al. developed an algorithm using MD simulation techniques to model CNT
bend junction structures formed by topological defects, i.e., pentagonheptagon
pairs. These three methods could give accurate results, but intensive computations
for the Schordinger equation and the particle dynamics of Newtons second law are
needed to predict the motion of nanotube materials. Thus it is difficult to make
quantum mechanics simulations.
On the other hand, simplified continuum and multiscale approaches seem more
practical for nanostructured smart materials. The modeling method must have an
ability to consider not only exceptional DC electrical transport but also AC behavior.
AC characteristics of nanotube materials are very important features for modeling
smart materials. P.M. Ajayan et al. reported that electrical transport in carbon nan-
otubes depend on frequency. The results explained an intrinsic resonance at a fixed
ultrasonic frequency of 37.6 kHz in the AC impedance spectra of nanotubes that
did not depend on the types or lengths of nanotubes. And until reaching an 8-MHz
frequency, the overall impedance showed a negative capacitance due to metal-tube
connections. P.J. Burke used Luttinger liquid theory for modeling the gigahertz
electrical properties and designed an RF circuit model of carbon nanotubes.
134
In
addition to the conductance of SWCNTs and MWCNTs, the conductance of bundles
(ropes) and networks (mats) of ropes in varying directions have to be determined
both for metallic and nonmetallic behavior. A.B. Kaiser et al. reported that the peak
in the frequency-dependent conductivity of single-wall carbon nanotube networks
is consistent with metallic conduction interrupted by nonmetallic defects that act as
barriers.
135
They further explored this model of metallic conduction interrupted by
barriers and two additional properties of SWCNT networks: frequency dependent
conductivity and nonohmic conduction. D. Qian and Y. Liu are developing multiscale
methods for the analysis of nanocomposite materials
136138
that may be applied to
nanoscale intelligent materials. Overall there has been little modeling of the elec-
tromechanical actuation properties of carbon nanotubes. The elastic properties and
the electrode kinetics at the nanotubeSPE interface are critical and have not been
modeled explicitly.
15.5.2 NANOTUBE STRAIN SENSORS
CNTs have been used as better alternatives to the conventional silicon tips in atomic
force microscopes because CNTs have flexible and mechanically robust properties,
and also the same repeated properties in air, vacuum, and water. Another enlarging
research field for CNT sensors is chemical force microscopy (CFM), which can
investigate the biological structure, for example, of a DNA sequence. Applying
electrostatic voltages can deform nanotubes. P. Poncharal reported that the charge is
located at the tip of a nanotube, and deflection is proportional to where is the
static potential. It was also found that the resonance frequency excited by alternating
the applied potential changes if masses from the picogram to femtogram level are
attached to a CNT. This property can be used for a nanoscale smart mass sensor.
173
As sensor systems are becoming more complex with many sensor nodes, a wiring
network is another issue to be resolved in using nanotube sensors in structural health
monitoring systems. Keat Ghee Ong et al. first discussed the possibility of passive
wireless communication of nanotube sensors.
175176
Such sensors are based on induc-
tor-capacitor resonant-circuits (LCs), which can be remotely monitored through loop
antenna. This passive wireless sensor enables the long-time monitoring of sensors
without batteries. Other new technologies such as wireless-LAN or Bluetooth have
the potential to be used as active wireless sensors for structural health monitoring
systems.
15.5.3 ACTUATORS BASED ON NANOSCALE MATERIALS
Researchers in the field of smart structures have been trying to overcome the limi-
tation of small strains or small forces produced by smart materials. Dynamic systems
are becoming more complex, and higher actuation efficiency is needed. In order to
resolve this situation, locomotion morphology and the physiology of animals
51
and
nanotechnology are topics of study for scientists and engineers. With the help of
bionanotechnology, biomimetic materials such as nanotubes for smart actuators have
considerable potential in nanoelectromechanical systems (NEMS). These actuators
provide greater work per cycle than previous actuator technologies and may have
better mechanical strength. In addition, CNTs offer high thermal stability, and the
actuators can be used in high-temperature environments.
15.5.3.1 Carbon Nanotube Electrochemical Actuators
The first actuator made of CNT was a macroscale sheet of nanotubes termed buckypaper
by Ray Baughman et al.
5759
This actuator used the change in dimension of the
nanotube in the covalently bonded direction caused by an applied electric potential.
The charge injection led to strain of the nanotube paper causing the assembly to
V
s
2
, V
s
2
bend. This excess charge was compensated at the nanotubeelectrolyte interface by
electrolyte ions forming the double layer. Soon after the buckypaper actuator, their
team demonstrated actuation of a single SWCNT in which the SWCNT was sus-
pended over a trench, and its edges fixed at the surface with a metal layer.
54
The
device was placed into an electrolyte, and a potential was applied to the SWCNT.
The change in SWCNT length was monitored by an AFM tip. These experiments
demonstrate that it is possible to use SWCNTs as actuators for driving pumps or
controlling nanofluidic valves.
Compared to the best known ferroelectric, electrostrictive, and magnetostrictive
materials, the very low driving voltage for CNT actuation is a major advantage for
various applications such as smart structures, multilink catheters, micropumps, flaps
for a microflying object (MFO), molecular motors, or nanorobots. Another advantage
is direct conversion of electrical energy to mechanical energy and high actuation
strain, strength, and elastic modulus. Since Faradaic actuators basically come from
the chargedischargecharge method such as a battery, it is possible to design a self-
powered actuator that can actuate motion based on stored charge in the CNT struc-
tural material without an external power source. However, since these CNT actuators
work in electrolyte, the bandwidth of an actuator is low and applications thus far
have been focused on artificial muscle.
7481
On the other hand, Von Klitzings group
in Germany used a solid electrolyte such as a PVA/Nafion/ H
3
PO
4
mixture for a
harder actuator and a wide bandwidth.
61
NASAs Jet Propulsion Laboratory in
California tried to develop bimorph actuators and force sensors based on carbon
nanotubes. This idea also came from double-layer charge injection that causes bond
expansion. These devices could generate and control displacements and forces on
a molecular scale, and can be embedded in circuits as NEMS devices.
15.5.3.2 Thermally Activated Actuators
The Nanoscale Science Research Group of the University of North Carolina dem-
onstrated that a double-layered cantilever on a MWCNT changed its curvature under
a temperature change. This is called a thermally actuated mobile system (TAMS).
63
The bilayer cantilever TAMS is an actuator/sensor that has two layers with different
thermal stress or intrinsic stress between the two materials. They reported that
greatest bending occurred at a thickness ratio of 1: 7 (thickness of aluminum/ thick-
ness of MWNT tAL/tMWNT), and the TAMS actuator was controlled by laser
heating.
15.5.3.3 Piezoelectric and Nanotweezer Actuators
The piezoelectric effect and high strength of BNT have the potential to be used in
developing high bandwidth structural actuators. Most of the papers investigating
the piezoelectricity of nanotubes are based on quantum mechanics, and there is
little experimentation. Controlling the chirality and producing aligned BNTs for
practical applications is still a problem. Eventually this synthesis problem will be
solved and BNTs may be widely used in smart structures.
183
Different nanotweezers
were developed using optical trapping. The optical system is complex, large, and
expensive, but if SWCNTs are used for nanotweezers, they will provide an advan-
tage in the development of various sizes and forms of nanotweezers.
183184
This
CNT nanotweezer has the ability to grab and manipulate nanosize objects in three
dimensions and may find applications related to nanomechanical structures.
15.5.3.4 Shape Memory Alloy and Platinum
Nanoscale Actuators
Shape-memory alloys (SMA) are already widely used in smart structures and arti-
ficial muscles.
203207
Notably nickeltitanium and copper-based alloys are available
and used to develop devices that can perform a wide variety of functions: as actuators,
components for smart structures, medical and orthodontic devices, safety valves,
eye glass frames, etc. But the actuation is indirect and requires high power to convert
electrical energy to thermal energy to cause actuation. The SMA actuators are being
formed as thin films to increase the cooling rate. Recently Weissmuller et al.
reported that an electrolyte-filled assembly of pressure-compacted platinum nano-
particles could be used as a smart actuator.
185186
They observed actuation strain up
to 0.15% for the platinum electrode, which is the same as for commercial ferroelec-
tric ceramics, but the applied voltage is small, similar to that of carbon nanotube
electro-chemical actuator cells. They also said that the strain depends linearly on the
potential, with little or no hysteresis.
15.5.3.5 Biological Molecular Actuators
Biological molecules are other possible types of nanoscale actuator. Three types of
cytoplasmic motors myosins that move on actin filaments, dyneins, and kinesins
that use microtubules as tracks are amazing biological machines that can be used
with nanotube materials as sensors and actuators. These examples of extreme
nanoengineering are fueled by ATP and convert its chemical energy into mechanical
energy. For example, a carbon nanotube can serve as a casing material for an active
motor or functionalized nanotube to be used as a molecular shuttle on a track. These
kinds of structures of nanotube sensors and actuators are gaining the spotlight as in
vivo bionanotube structures.
187188
Compared with protein structures, DNA is another
new and up-and-coming smart actuator since it is small, simple, and its function is
well understood. The NASA Institute for Advanced Concepts has already started
research to develop a DNA actuator with SWCNTs. Several viral protein linear
(VPL) actuators were placed in parallel and series to multiply force and displace-
ment, respectively, and a SWCNT-outfitted VPL actuator was shown
190
to open and
close depending on the pH level of the environment. Basically this change in viral
protein is required for the process of membrane fusion; i.e., the fusion of viral and
cellular membranes is essential for infection of the cell.
Other biological molecules being developed for actuation uses inlude proteins.
Changing the pH level causes the protein molecules to unfold, producing an
actuation force. The force may be used for morphing structures and to form active
membranes.
15.6 MANUFACTURING OF CARBON NANOTUBE
AND NANOFIBER INTELLIGENT MATERIALS
Manufacturing of carbon nanotube and nanofiber intelligent materials involves
the following steps: (1) nanotube synthesis and purification, (2) functionalization
of the nanotubes, (3) dispersion and casting of the nanotubes or nanofibers in a
structural polymer electrolyte (SPE), which includes forming the structural mate-
rial, electroding, and encapsulating the material, and (4) testing and character-
ization of the material performance. These steps are discussed in the following
sections.
15.6.1 SYNTHESIS OF NANOTUBES
The SWCNT, MWCNT, and CNF materials can be used separately and in combi-
nation to make force actuators, strain sensors, and power generators. The actuation
capability and cost can be tailored for a particular application. The synthesis step is
important to improve the size and actuation properties of nanotube materials. The
complete processing of smart materials starting from nanotube synthesis to final
device fabrication is performed at UC. Nanotube synthesis is accomplished using a
commercial EasyTube nanofurnace, as discussed in Chapter 5.
31
Objectives of
nanotube synthesis experiments include growing long CNTs on or between sub-
strates to improve the load transfer from the nanotube to the matrix and the trans-
duction properties of CNT hybrid materials. During synthesis, MWCNTs have
sufficient rigidity to overcome the gravity, buoyancy, and van der Waals forces to
keep growing in the same direction up to a point, and then they bend. The MWCNT
arrays grow at about 5 m/min and millimeter lengths may be possible. The length
of the nanotube is also controlled by the covering or depletion of the catalyst.
An array actuator using MWCNTs grown on substrates is being developed to
overcome the problems of alignment, ion exchange, and load transfer for electro-
chemical actuation. MWCNT arrays that are 250 square by 130 tall blocks were
grown on a silicon substrate by FirstNano, Inc. using the EasyTube nanofurnace.
31
Also, arrays of 2.5 mm long SWCNTs were grown Japans Agency of Industrial
Science and Technology
208
using the nanofurnace with the bubbler attachment. Water
vapor was used to react with amorphous carbon to prevent the catalyst from being
covered and becoming inactive, e.g., H
2
O C
am
oCOH
2
. This is a large advance
in the development of smart materials because longer nanotubes will allow easier
processing and provide greater actuation and material strength.
A key component of synthesis research is to couple the synthesis of the
nanotubes with detailed atomic scale characterization of the structure, composi-
tion, and morphology of the tubes and to increase the lengths of the nanotubes.
The structure, diameter, uniformity, and chirality of the MWCNT and the structure
of crystallites formed by self-organization of the MWCNT can be determined by
environmental scanning electron microscopy (ESEM), high-resolution transmission
electron microscopy (HRTEM), atomic force microscopy (AFM), micro-Raman,
and electron and diffraction techniques. ESEM is very useful for characterizing
nanotube growth and nanocomposite failure surfaces without a need for much
preparation of the material. HRTEM is useful to study the wall structure of
SWCNTs and MWCNTs. Raman is used to determine the nanotube diameter.
The diameter is related to the (m, n) indices, and the chirality of SWCNTs can
be determined by knowing the diameter based on the Raman results.
15.6.2 FUNCTIONALIZATION OF NANOTUBES
Purification and functionalization are essential in defining active surface area,
charge transfer rate, and adsorption/desorption at the CNT surface. Creating a
strong bond (interface) between the nanotubes themselves and between nano-
tubes and the matrix materials in the case of composite materials is crucial for
design of sensors and actuators. UC has developed a plasma polymerization
process to coat or functionalize CNTs with different materials.
159164
In a plasma
gas environment, the surface atoms of the CNT become activated. At these active
centers, a polymer film will form on the CNT surface. The polymerization
conditions can be controlled so that the structure of the coating film has the
desired number of functional groups. The plasma coating process is well described
in Chapter 8.
Functionalization research related to smart materials is focusing on increasing
the electrochemical ion exchange at the nanotube surface when the nanotubes are
cast into a solid polymer electrolyte (SPE). The individual nanotubes are coated
with an ultrathin coating of acrylic acid or polypyrrole using plasma polymerization
before dispersing and mixing in the polymer electrolyte using simultaneous sonica-
tion and shear mixing. The different functionalization methods can be tailored to
change the mechanical, electrical, and magnetic properties of the CNT due to a
combination of dimensional changes and interface properties. In combination with
plasma coating, surfactants used with ultrasonication appear to be a good approach
to disperse nanotubes. Using 1% by weight of random SWCNTs or nano-fibers and
the UC functionalization and dispersion methods, 30 to 40 wt% increases in elastic
modulus of epoxy have been achieved. Up to 15% by weight of nanotubes was
incorporated in epoxy, but the dispersion techniques for this large loading are still
being developed. Nafion has also been used to coat CNTs to make electrochemical
actuators. The experience in reinforcing composites is useful for developing CNT
actuators. The challenge is to improve the speed of the ion exchange and increase
the strain transfer of the nanotube actuator in the SPE. Sometimes the nanotubes
may be filled with the SPE. A nanotube filled with the SPE is called a composite
nanotube, because the nanotube is the host material, while integrating nanotubes in
the SPE produces a nanotube composite because the SPE matrix is the host material.
15.6.3 CASTING NANOTUBES AND NANOFIBERS IN STRUCTURAL
POLYMER ELECTROLYTE
There are particular challenges that must be met before nanotube properties
can be fully exploited. The main problem is to attain good dispersion and
interface bonding between the CNT and the polymer matrix and achieve good
load transfer from the matrix to the CNT during loading, since colloidal mate-
rials such as CNTs do not spontaneously suspend in polymers.
192199
Dispersion
may be the most critical part of integrating any of the types of nanotubes into
polymers or other host materials. Sonication is often used to disperse nanotubes
in solvents or resins. Inadequate sonication power level will not disperse the
nanotubes satisfactorily. On the other hand, an excessive power level will break
the nanotubes apart and shorten the nanotubes. During sonication, resonant
bubbles are formed and collapse, generating very high local pressures and
temperatures. Different sizes of nanotubes require different levels of sonication
to achieve dispersion without shortening the nanotubes. To resolve these issues,
many groups are trying to disperse CNTs homogeneously and uniformly in
matrix materials. Solvents including amides, NN-dimethylformamide (DMF),
and N-methylpyrrolidone (NMP) were reported to provide solubility of CNTs
for different applications. Using appropriate surfactants, CNTs can also be
solubilized in water.
90
Also, SWCNTs have been solubilized by functionalizing
the end caps with long aliphatic amines or the side walls with fluorine and
alkanes. By using an ultrasound mixer or using strong acids, good CNT dis-
persion can be achieved
68
for lower percentages of nanotubes from about 1 to
3 wt% of SWCNT. Dispersing larger percentages of nanotubes in polymers
remains a problem. Recently NaDDBS has been reported to be an excellent
surfactant for dispersing nanotubes in water.
90
A filtering and drying procedure
is needed to prepare the surfactant-dispersed nanotubes for incorporation into
epoxy and PMMA polymers. High shear mixing is also critical to separate
nanotubes.
For the smart structures applications requiring the high conductivity of CNTs,
coating of metals and organic conductive polymers (CPs) on the surface of CNTs
to produce CNT-reinforced composites is another method.
192199
CNTs coated with
polypyrrole, Ni, Co, Ti, W, Pd, Au, Al, and Fe to form composites are not uniform
and homogeneous, resulting in tangling and random distributions.
125
CPs such as
polypyrrole, polyaniline, and polythiothene can be used as the conducting matrices,
but these are mechanically weak and have to be oxidized and doped by a counter
anion to achieve significant conductivity.
126,128,129
With large enough percentages of
CNT in the insulating matrix, the nanotube-reinforced composite may be an ionically
conductive structure that has satisfactory mechanical strength. Careful consideration
to choosing the appropriate solid polymer matrix and coating materials is a key issue
in synthesizing smart materials using nanotubes. Loading polymers with nanotubes
can be done using SWCNTs, MWCNTs, or CNFs. SWCNTs that have perfect atomic
structures are electrically the best fillers, but the cost is high and the chiralities
cannot be easily separated. MWCNTs do not have good shear stress transfer
between the individual shells, and the number of walls usually increases the number
of defects. It appears that the inner layers provide mechanical support, unless
electrical contact is made with the inner layers. CNFs have lower electrochemical
properties than MWCNTs. At UC, the ultrasonication and shear mixing steps are
combined in the procedure for the dispersion of nanotubes in epoxy. This shortens
the time of sample preparation. The combined setup is shown in Fig. 15.2. A proce-
dure for processing smart nanocomposite materials now in development includes (1)
nanotube synthesis, (2) plasma polymerization or acid oxidization functionalization,
(3) dispersion in a solvent, annealing, and then dispersing in the resin, (4) casting
using high-pressure consolidation or vacuum degassing, (5) curing, and (6) cutting,
electroding, and encapsulating the smart material.
Precise control and optimization of all processing steps are extremely impor-
tant to fabricate nanostructured smart materials. Therefore we are developing a
vacuum processing system for the fabrication of CNT smart materials. The vacuum
processing system shown in Fig. 15.2 allows sonication and mechanical mixing
of the nanotubes and SPE at elevated temperature in a partial vacuum or inert gas.
High-pressure consolidation of the liquid nanocomposite material to fill the larger
open nanotubes with the resin is also being done. This fabrication procedure along
with using longer nanotubes is expected to significantly improve the performance
of nanocomposite materials. The actuation and strength properties of the nano-
composites produced are determined using electrochemical impedance spectros-
copy, load and displacement testing, mechanical strength testing, and microscopy.
The following section discusses recent research and presents ideas to develop
nanotube sensors and actuators for smart structures and nanotube-based health
monitoring systems.
45,6668,200202
15.6.4 CARBON NANOTUBE COMPOSITE STRAIN SENSORS
Carbon nanotube piezoresistive sensors are being considered for self-monitoring
of composite structures. The CNT and CNT-epoxy have a number of different
sensing properties. There has been investigation of these properties and what
properties are optimal for different applications. Figure 15.3(a) shows a a2-mm
thick CNT-epoxy sensor with 5% randomly oriented CNT. The piezoresistive effect
(resistance change with pressure) is also shown. The contact resistance
3543
of the
CNT-epoxy/metal contacts causes the large slope in Fig. 15.3(a). Silver-filled
epoxy can eliminate most of the high-contact resistance. Resistance that decreases
with compressive loading may also be due to pushing the CNTs closer together
improving hopping/tunneling and direct contact in the epoxy matrix. The electros-
trictive effect (quadratic charge due to strain) of the matrix material is shown in
FIGURE 15.2 Simultaneous ultrasonication and shear mixing (a) of a nanotube polymer in
a glove box (b).
Humidity, temperature,
pressure and
atmospheric control
Ultrasonicator
tip and rotary
mixer
(a) (b)
Fig. 15.3(b). A 0.5-volt peak signal is produced due to an impact; Fig. 15.3(c)
shows a 100% CNT film sensor, R 12 ohms; Fig. 15.3(d) shows a pressure-cast
CNT-epoxy barrel sensor, R 400 ohms; Fig. 15.3(e) shows a CNT-epoxy sensor
plate with electrodes on top and bottom. These are early examples of self-sensing
structural materials.
Since macroscale strain sensors are needed, the nanotube resistance properties
must be brought to the microscale. Different nanotube systems including individual
SWCNTs, MWCNTs, CNFs and ropes and mats can be used for making piezore-
sistive sensors. It may also be possible to build nanotube switches sensitive to
very small pressure changes. A carbon nanotube neuron crack sensor for SHM is
shown in Fig. 15.4. The crack sensor can be a highly reliable safe-life sensor that will
ensure that no crack length larger than the spacing between two parallel carbon
nanotube sensor nerves exists. Modeling and different polymers are being used to
improve the sensitivity. NASA Langley researchers
200
are also developing a carbon
nanotube strain sensor. Rice University researchers
177
have developed a buckypa-
per piezoresistive sensor.
FIGURE 15.3 SWCNT sensors built at University of Cincinnati: (a) 5% CNT load sensor
and resistance versus compressive load, (b) voltage (0.5 V peak) versus time due to impact
on an electrostrictive polymer, (c) 100% SWCNT film sensor, R 12 ohms, (d) pressure-cast
CNTepoxy barrel sensor, R 400 ohms, (e) CNTepoxy sensor plate with electrodes on top.
0
20
40
60
80
100
0 50 100 150 200
Load (N)
R
e
s
i
s
t
a
n
c
e

(
K
-
O
h
m
)
(a)
(d)
(e)
(c)
(b)
2 2 1.5 1.5 2.5
10
1
0.5
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5 0 1 1
Time (sec)
V
o
l
t
a
g
e
Impact response of CNT/epoxy
15.6.5 CARBON NANOTUBE-BASED BIOSENSORS
Biomedical applications of nanoscale intelligent materials have an especially large
potential. Three applications can be categorized: (1) biomedical diagnostic tech-
niques, (2) drugs, and (3) prostheses and implants. Biomedical applications for use
outside the body, including diagnostic sensors such as DNA, RNA chips, lab-on-a-
chip, and micro-total analysis systems (TAS), are suitable for analyzing biomolec-
ular samples. For internal use, several researchers are developing anticancer drugs,
bio-MEMS devices, and gene therapy. Other researchers are working on prostheses
and implants that include nanostructured materials.
Current biological sensors commonly rely on optical detection principles that are
inherently complex, requiring multiple steps, multiple reagents, preparative steps,
signal amplification, and relatively large sample sizes. Therefore, these techniques are
highly sensitive and specific, but difficult to miniaturize. In order to overcome these
limitations, a carbon nanotube array-based sensor is considered using the electro-
chemical properties of nanotubes for biosensors. Based on this idea, different possible
uses of nanotubes for sensors were studied. Types include strain sensors, electrochem-
ical biosensors, gas chemical sensors, temperature, and light sensors. A limiting factor
for developing artificial organs such as eyes or ears is the interface between the
biological component and an artificial silicon chip. A remarkable solution for this
problem is a CNT array that can directly contact a cell in vivo, since CNTs provide
highly biocompatible surfaces. One example is a vision chip consisting of an array
FIGURE 15.4 A CNT neuron continuous strain sensor and its dynamic strain response after
low-pass filtering (cutoff freq 30 Hz, 60dB gain): (a) MWNT/PMM 10 wt% neuron (300 mm u
5 mm, 10 K resistance) on a glass fiber plate; (b) response due to an initial tip end displacement
of 1 cm; and (c) response due to an impact at the tip of the beam. A laser sensor is used to
measure the displacement of the beam for reference. (a) Composite Actuator (b) Linear Actuator
(c) Tripod Actuator .
(a)
Strain sensor
Laser sensor
Time (100 ms/div)
V
o
l
t
a
g
e

(
1
0
0

m
V
/
d
i
v
)
(b)
Strain sensor
Laser sensor
Time (50 ms/div)
V
o
l
t
a
g
e

(
1
0
0

m
V
/
d
i
v
)

(c)
v
of electrically conductive CNT towers grown directly on the surface of a silicon chip.
A group has already tested the implant materials with retinal ganglion cells in vitro
and proved the excellent biocompatibility.
201
The nanotube-based electrochemical sensor can reduce the drawbacks of other
electrode materials that include short life, low conductivity, and low reliability. For
biosensing applications, carbon nanotubes have advantages of small size, large surface
area, high sensitivity, fast response, and good reversibility at room temperature. At
the same time, carbon nanotube biosensors can be integrated with microelectronics
and microfluids systems to gain advantages in miniaturization, multiplexing, and auto-
mation. Carbon nanotubes functionalized with special chemical groups can form
electrochemical biosensors for enzyme immobilization and stabilizers and mediators
used in cancer research. UC is focusing on the development of a carbon nanotube
array-based cancer sensor. The nanotube array becomes a miniature bioelectronic
sensor with a large number of functionalized antibody sites for high sensitivity. The
design of the nanotube array determines the sensitivity of the sensor. The research
involves synthesizing MWCNTs in an interdigitated pattern, as shown in Fig. 15.5.
Based on this array pattern design, the fabrication and processing can be done using
semiconductor processing techniques. The complex CNT array pattern is expected to
improve the detectability and fast response of the sensor and provide reliable data
with minimal false positive and false negative readings. This array could also be put
on a chip in a portable device to create a lab-on-a-chip system with a microflow
channel.
The semiconductor processing steps that form the array are shown in Fig. 15.6.
These steps are preliminary and are under early development and can be used to grow
large planar arrays of MWCNTs that can be sectioned to provide many sensor elements.
The critical step to develop the CNT array-based sensor is the fabrication process.
Figure 15.6 illustrates the proposed detailed sequential steps in the fabrication process
of the sensor segment. The starting substrate in Step 1 is either a highly resistive Si
single crystal wafer or sapphire that provides an insulating base, which is required for
the wiring of the segment. Step 2 involves deposition of highly conductive (doped)
poly Si for planting the CNT catalyst on it. Doped Si provides good electrical contact
between the CNT and the substrate. The CVD technique is used for depositing poly
Si doped either with P or As. As an alternative solution, one can use an insulating Si
wafer that is ion-implanted on the CNT growth side for increasing the electrical
conductivity. Steps 3 to 7 involve patterning and require sputtering of a thin metal film
FIGURE 15.5 Concept of a nanotube interdigitated array biosensor.
Vertical double
row MWCNTS
functionalized
with antibodies
+ electrode
electrode
(2040 nm) of catalyst such as Co and thermal annealing of the film in hydrogen at
825qC. During the annealing, catalytic semispheres of Co islands are formed, driven
by surface tension to lower the total energy. Simultaneously a chemical reaction
between the metal and the Si substrate takes place, producing highly conductive
CoSi
x
. This compound is expected to provide a strong mechanical bond between
the CNT and the substrate and also serve as a low resistivity electrical junction.
The patterned substrate with catalyst provides aligned growth of MWCNTs during
Step 8. Functionalization of the array by solution methods or plasma coating is
shown in Step 9. The patterning and functionalization will be optimized to match
the specific antibodies to be immobilized and the growth factors and cytokine to
be sensed. Medical doctors at UC are involved in the design. Immobilization of
the antibodies is shown in Step 10. In this design, the electrical signal can be
provided to each individual column of aligned carbon nanotubes or in almost any
arrangement desired to tune the sensitivity.
The nanotube sensor array can be patterned to have a desired connectivity of the
individual pairs of nanotube sensors shown in Step 10 of Fig. 15.6. The highest
sensitivity of the array can be examined by modeling the impedance of the whole array
when the antibody-growth factor coupling occurs. The nanotube spacing in the array
and the nanotube diameter, length, and functionalization must be optimized to allow
the greatest sensitivity while rejecting small changes in sensitivity that could lead to
false positive and false negative indications. The functionalization improves the elec-
trochemical response of the array to specific proteins, enzymes, and growth factors on
the surface of CNTs. At UC, various chemical groups such as COOH and NH
2
are
coated on the nanotubes to optimize the electrochemical response from the CNT array-
based electrode acting as a host for different biospecies. The functionalization can be
done in two ways for comparison. One is a reagent bath in which acids are used to
FIGURE 15.6 Fabrication steps for the double-row interdigitated nanotube array sensor
shown in Fig. 15.5.
High
resistivity Si
Doped poly Si
Photoresist
UV
Shadow mask
Co or Ni
Co or Ni
Sputtering of Co
Etching
Coat with photoresist
Deposition of doped poly Si
Step 1
Step 7
Step 8
Step 9
Step 10
Step 2
Step 3
Step 4
Step 5
Step 6
Thermal treatment
at 825C in H
o
Co
CoSi
x
CNT
CNT
CNT growth
Functionalization
Immobilization
Antibodies
attach specific function groups to the nanotubes. This is the current conventional
method that is widely used to functionalize nanotubes. Plasma surface modification is
the other very promising method for nanotube functionalization. In the plasma tech-
nique, for example, water and oxygen may be used to functionalize the surface of the
MWCNT array instead of using a monomer such as styrene or pyrrole that is used for
polymer coating. In the plasma phase, the carbon atoms on the surface are active and
react with water or oxygen molecules to form COOH and OH groups. If NH
2
is
needed, ammonium hydroxide and NH
3
gas can be used as the monomer as well.
The carboxyl function groups functionalized on the nanotube surface will react
with biospecies such as enzymes, antibodies, and oligonucleotides and form strong
covalent bonds to attach biospecies to the surface of the nanotube. Depending on the
different biospecies attached on the surface, the aligned CNT can be used to detect
DNA, hydrogen peroxide, glucose, and other biomolecules. As an example, a carboxyl
function group was attached to a MWCNT by surface modification. An amino function
group on the oligonucleotide probe can react with the carboxyl function group to form
covalent bonds and attach the probe on the surface of the nanotube. When this kind
of biosensor is put into the test solution, the species in solution can attach on the probe.
When immobilization occurs, a different electrochemical response occurs; the change
in the impedance spectra or current is used to detect the presence and the amount of
the complementary sequence. Figure 15.7 shows the experimental setup for the elec-
trochemical bionanosensor for on-site analysis. The electrochemical system is based
on amperometric or electrochemical impedance measurements.
Once the nanotubes are functionalized using the plasma system, immobilization
of antibodies on the nanotubes must be done. Antibody immobilization formats can
be broadly categorized into two classes: direct labeling experiments and dual antibody
sandwich assays. These methods are possible for use with the nanotubes. However,
based on our experience functionalizing carbon nanotubes, the carboxyl function group
is very active and it can react with the NH
2
function group via carbodiimide activation.
A possible approach is to use the plasma coating method to functionalize the nanotubes
based on our existing technique, use solution methods to immobilize the antibodies,
and use techniques such as FTIR, and TOFSIMS to characterize the immobilization.
FIGURE 15.7 Electrochemical analysis setup using (a) a Gamry Potentiostat, and (b) func-
tionalized CNF (ESEM image by Srinivas Subramaniam, University of Cincinnati Materials
Characterization Center.)
CNF-GO
x
electrode
Buffer sol.
SCE
Counter
Working
Potentiostat
PCI board
Magnetic stirrer
The amperometric method, especially cyclic voltammetry, is used because it can
measure the concentrate-dependent current through the CNT array-based electrode
coated with a specific chemical group. Each human has on the order of 50,000 genes
that may play a critical health-related role. Biosensors that can perform genetic analy-
sis, therefore, have tremendous potential. At UC, initial experiments to develop a
nanotube biosensor were performed.
Figure 15.8 shows electrochemical impedance spectroscopy (EIS) data (Nyquist
plot) with imaginary (capacitive) versus real (resistive) components of the complex
impedance. Three different potentials were tested: 0.2 V, 0.4 V, and 0.6 V. Figure
15.8(a) shows the EIS results before the immobilization of GO
x
on the surface of
the CNF film and Figure 15.8(b) shows the EIS results after GO
x
was immobilized on
the surface of the CNF electrode. With the immobilization of the GO
x
on the nanotubes,
the impedance of the sensor changed as shown by comparison of the responses (a)
to (c) in Fig. 15.8(a) and Fig. 15.8(b). The change in impedance is due to the change
in the electrical resistance and capacitance of the electrode due to the immobilization
of the GO
x
. This result indicates the potential to use the bioelectronic sensor for
label-free sensing. Figure 15.8(c) shows the amperometric response of the biosensor
in which 1 mM glucose is added at each step at a potential of 0.5 V in PBS (pH 7).
Even though there is slight noise in the signal caused by the 1-PA resolution level
of the Gamry system, the current response increased immediately after each addition
of 1 mM of glucose in the PBS solution. This change in response was repeated five
times and retained the same value for 200 s. This high sensitivity and stability are
advantages of CNF-Go
x
biosensors for commercial applications. The structure of
the CNF provides a large surface area to immobilize the GO
x
and maintain this high
activity. At the same time, CNFs could promote the electron transfer in the bioelec-
trochemical reaction.
A biosensor architecture based on carbon nanotubes may detect cancer and other
pathogens at the earliest stages using electrochemical sensing of molecular signatures
FIGURE 15.8 Biosensor characterization using glucose. (a) Electrochemical impedance
spectra for the biosensor at DC potentials 0.2 V (a), 0.4 V (b), and 0.6 V (c) before immo-
bilization of GO
x
; (b) after immobilization of GO
x
on the CNF; (c) amperometric response at
an applied potential of 5 V versus the SCE reference electrode from the CNF-based sensor
due to the successive addition of 1 mM glucose in a 0.05 M phosphate buffer solution (pH 7).
2000
1500
1000
0
0
500
Z
i
m
a
g

(
O
h
m
)
Z
real
(Ohm)
500 1000 1500 2000
(a)
(b)
(c)
2000
1500
1000
0
0
500
Z
i
m
a
g

(
O
h
m
)
Z
real
(Ohm)
500 1000 1500 2000
(a)
(b)
(c)
3
3.5
2.5
2
1.5
1
10
6
0 200 400 600 800
C
u
r
r
e
n
t

(
A
)
Time (sec)
(a) (b) (c)
Increasing
freq.
that can identify diseases. The nanotubes are good biosensors because their small
size can allow detection of small numbers of biomolecules and the nanotube resis-
tance and capacitance are sensitive to the attachment of biomolecules on the walls
of the nanotube. The smaller size and higher sensitivity of the sensor may provide
early detection and better outcomes for many diseases including cancer. The
approach using interdigitated array architecture and a plasma coating method is
expected to improve sensitivity and speed of detection as compared to existing types
of sensors, and as compared to using a multiplexing approach with statistical analysis
of separated individual arrays.
15.6.6 CARBON NANOTUBE AND NANOFIBER HYBRID ACTUATORS
In order to develop actuators for practical structural applications, there are many
challenges. Therefore at UC we have been developing a polymer host for nanotubes
that can provide ion exchange for the electrochemical effect.
192199
Little or no
previous research has been done using CNF for electrochemical actuation. Even
though the electrochemical properties of CNFs are fewer, we are focusing on CNFs
for new actuators because the cost is low at $0.25/gm (a$100/lb), and incorporation
of the nanofibers into polymers is easier because the fibers are large (a70150 nm).
The CNF material is also readily available in commercial quantities and large
amounts can be made. The Bionanotechnology Lab at UC is developing for the first
time a carbon nanofiber (CNF)polymethylmethacrylate (PMMA) composite mate-
rial that has electrochemical actuation properties. A combination of solvent casting
and melt mixing was used to disperse CNF in PMMA and cast thin films of the
material. Solution casting of the actuator material is shown in Fig. 15.9(a) for use
as a wet actuator. Since PMMACNF composites have electrical resistance, sput-
tering of metal particles on the surface of PMMA/CNF composite is one approach
being investigated to develop the activator and provide a uniform voltage field.
FIGURE 15.9 Manufacturing steps for the CNFPMMA hybrid actuator. (a) CNF in a
polymer matrix; (b) actuator strips as produced and with sputtered gold electrodes; (c) SEM
image of fracture surface with fiber pull-out. (SEM image by Robert Gilliland).
Nanocomposite
Nanofiber
Polymer
+
CNT/PMMA
composite
Insulated
layer
Gold
(a) (b) (c)
The electrodes of the actuator should be insulated from the electrolyte because
we have found that only the top part of the actuator provides most of the actuation.
This is due to shunting through the electrolyte and a nonuniform voltage along the
actuator. Tailoring of the voltage along the actuator and encapsulation of a partially
hydrated actuator are approaches that are under investigation to overcome this
problem and to further improve the actuation performance. The CNFPMMA
material produced using the solution casting procedure is shown in Fig. 15.9(b). The
actuators shown are from 1 to 6 in. in length. Larger actuators and components can
be made because of the low cost of the CNFs and chemicals. Fig. 15.9(c) shows the
failure surface. The nanofibers are shown pulling out of the matrix material. The
interfacial bonding of the CNF to the polymer is currently weak and will be improved
using the plasma coating and dispersion methods.
Solution casting of the CNFPMMA actuator for use in wet electrolytes is shown
in Fig. 15.10(a). Figure 15.10(b) shows the experimental setup for the CNFPMMA
actuator. These aqueous-based actuators are immersed in a 2 M NaCl solution. Square
FIGURE 15.10 Solution casting and testing of the wet CNFPMMA hybrid actuator:
(a) solution casting; (b) typical experimental setup; (c) actuator response; (d) measured
response near fixed end (out of water) of actuator.
(c) (d)
(a)
4 6 8 10 12 14
0
2
4
6
8
0.9
0.8
0.7
0.6
0.5
Displacement & applied voltage
Time (sec)
Low voltage (8 V) High voltage (20 V)
CNT/PMMA
solution
CNT/PMMA
solution
Polymer as
insulator
Laser displacement
sensor
Vision sys.
Op. amp &
NI board
Potentiostat
PCI board
Counter
SCE
Working
Electrolyte
CNT actuator
Drying
Drying
(b)
D
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
A
p
p
l
i
e
d

v
o
l
t
a
g
e

(
V
)
wave potentials are applied between the platinum counter electrode and the working
electrode. The displacement of CNFPMMA actuator is shown in Fig. 15.10(c) and is
measured by a laser displacement sensor (Keyence, LC-2400 Series) and monitored
using a vision system (Intel Pro PC Camera). Note the very large actuation at 20 volts.
This is due to surface actuation because of poor dispersion of the CNF. Also notice that
the bending occurs mostly near the top of the actuator. This is because of shunting in
the electrolyte and because the resistance of the actuator reduces the voltage in the
lower part of the actuator. Improved dispersion, processing, an insulated electrode, and
array methods will greatly improve the stiffness and bandwidth of the actuator. This is
a very exciting material because of the low cost and good actuation.
A square wave with amplitude of +8 V was applied at frequencies ranging from
0.2 to 10 Hz. The response was measured near the fixed end of the actuator out of
the water at 0.5 Hz and is shown in Fig. 15.10(d). The characterization of the actuators
is done using a specially built test system for electrochemical actuators. This PC-
based system can be built using readily available components for a total cost of about
$30,000. The displacement of the actuators is measured using a laser displacement
sensor (Keyence, LC-2400 Series). Square wave potentials are applied between two
electrodes using a National Instruments PCI board and an operational amplifier
designed and built to drive electrochemical actuators. The square wave excitation is
an effective way to drive the amplifier and actuator with various amplitudes and
frequencies using the PCI D/A board. In order to supply enough power to drive the
actuator, a voltage follower and noninverting amplification were designed using the
OPA2544 (TI) with a gain of 20: 1.
Solution casting of a dry actuator material is similar to that for the wet actuator,
Fig. 15.11(a). However, in the dry actuator procedure, the SPE contains the ion
exchange material. A dry actuator was constructed according to Fig. 15.11(a) using
individual CNF cast in a PMMA LiBF
4
SPE. A loading of a10% CNF is used in
the actuator in this example. The actuator is formed using two CNF actuators, one
on each side of tape. One actuator is used as ground and the other is switched
between 2 and 2 volts. The test setup is shown in Fig. 15.11(b), and the actuation
is shown in Fig. 15.11(c). Because the actuator is not hydrated, the actuation is slow
and the amplitude is small compared to the wet actuator. Some creep also occurs,
partly because the actuator is very flexible and the two actuators are not exactly the
same size. Individually coating the CNF, uniformly dispersing the CNF, pressure
casting, and aligning the CNF will improve the actuation. Allowing the material to
absorb a small amount of moisture would also improve actuation. There is an unusual
characteristic of this actuator. Note in the time response of the actuator shown in
Fig. 15.11(d) that each time the voltage switches, the actuator returns to equilibrium
quickly and then more slowly begins to actuate in the opposite direction. This is the
interplay between the elasticity and ion exchange in the solid polymer electrolyte.
A summary of the main properties of carbon nanotube and carbon nanofiber
hybrid actuators is given in Table 15.4. This is the first time some of these materials
have been characterized, and much more detailed characterization is planned.
We are developing the electrical equations of a CNT actuator in an electrolyte.
A 2 M NaCl electrolyte solution is used for the experiments. Details of the modeling
are given in References 6668. Electrochemical impedance spectroscopy (EIS) was
performed on three-electrode cells, with a dry CNF actuator as the working electrode,
a saturated Calomel reference electrode (Gamry, Part No. 930-03), and using plat-
inum as the counter electrode. EIS measurements were performed using a Gamry
Potentiostat (Model: PCI4/750) coupled with the EIS software (Gamry, EIS300).
The cell was equilibrated for several hours after each step. The electrochemical
impedance spectroscopy (EIS) results are shown in Fig. 15.12.
The EIS results and modeling using Randles circuit are discussed in References
67 and 68. Randles circuit is an equivalent representing each component at the
electrodeelectrolyte interface and in the solution during an electrochemical reaction.
The electrode kinetics model the change from electronic conductivity to ionic con-
ductivity, in the simplest way, in parallel as a double-layer capacitor and a polarization
resistance that models the electrode kinetics, and this parallel capacitor and resistor
combination is in series with a resistor that models the electrolyte solution resistance.
The CNT actuator must now be modeled as it charges and discharges under an
AC environment wherein the actuation is proportional to the basic charge transfer
rate. Although the electrolysis mechanism with an AC voltage input has very complex
characteristics, we can model this effect as an impedance element. The relationship
FIGURE 15.11 Solution casting and testing of the dry CNFPMMA hybrid actuator:
(a) solution casting; (b) test setup; (c) dry actuator with two-sided actuation; and (d) measured
response due to 0.5 Hz r 4 V excitation.
1.6
1.65
1.7
1.75
1.8
D
i
s
p
l
a
c
e
m
e
n
t

(
m
m
)
CNT/PMMA
solution
CNT/PMMA
solution
Drying
Drying
PMMA/
LiBF
4
solution
CNF
actuator
Vision
sys.
Laser
displacement
sensor
Operational
amp.
Ni
board
(a)
(c) (d)
(b)
20 25 15 30
4
2
0
2
4
A
p
p
l
y
i
n
g

v
o
l
t
a
g
e

(
V
)
Time (sec)
TABLE 15.4
Properties of Carbon Nanotube and Carbon Nanofiber Hybrid Actuators
Type of Actuator Wet Environment (NaCl Solution) Dry Environment (with SPE)
SWCNT buckypaper Fast, large strain, expensive, good
cohesion property to make
buckypaper, low strength
Slower response and smaller
actuation than wet
MWCNT buckypaper Good actuation strain and lower cost
than SWCNT
Slower response and smaller
actuation than wet
CNF buckypaper Low cost, difficult to make buckypaper,
low mechanical strength, needs higher
voltage and current to actuate
Slower response and requires larger
voltage, good mechanical loading
with SPE
CNF coated with
polypyrrole to form
bucky paper
Improved mechanical loading, similar
actuation properties as CNF
buckypaper
Slower response and requires larger
voltage, good mechanical loading
with SPE
SWCNT composite
made with PMMA
polymer
Fast, large strain, best among nanotube
composites, expensive
TBD
MWCNT composite
made with epoxy
polymer
Good strength and actuation, need to
laminate
TBD
CNF composite made
with epoxy polymer
Needs higher voltage and current, good
mechanical properties, dispersion
needs to be improved, has smart
structures applications
TBD
TBA = To be determined.
FIGURE 15.12 The EIS responses at cell potentials of 0.1 to 0.4 V: (a) 0.05 to 100 KHz;
and (b) enlarged view of high frequencies between 100 Hz and 100 KHz.
Increasing
frequency
3000
2000
1000
0
0 2000 4000 6000 8000
0.2 V
0.3 V
0.4 V
Z
i
m
a
g
(
O
h
m
)
Z
real
(Ohm) Z
real
(Ohm)
(a)
0.1 V
Z
i
m
a
g
(
O
h
m
)
150
100
50
0
100 150 200 250
0.2 V
0.3 V
0.4 V
0.1 V
(b)
between mechanical properties and electrical properties can be expressed approxi-
mately by the linear relation:
(15.1)
Here, H is the actuator strain, q is the charge per unit volume, V is stress, V is input
voltage, t is time, D is the strain-to-charge ratio, E is the elastic modulus, the double
layer capacitance is C
v
, and x and y are constants that depend on the construction
of the nanotube smart material and the type of electrolyte used. In this method, the
SPE will be designed to increase the capacitance and strain-to-charge ratio, while
maintaining the highest elastic modulus of the material as possible.
Based on a concept developed at UC, by monitoring the voltage source, a self-
sensing actuator can be formed as shown in Fig. 15.13. To implement this practical
self-sensing actuator, an electrical bridge circuit was used to measure strain, as shown
in Fig. 15.13(a). The electrolyte and CNT contact act as a resistance; R, the interface
between the polymer and the electrolyte is modeled as the double layer capacitance,
C
v
, and the diffusion element is Z
D
. The self-sensing actuator will be used in the
feedback control system shown in Fig. 15.13(b) that allows the scheme to take advan-
tage of the guaranteed stable collocated output feedback control. Compared to the best
known ferroelectric, electrostrictive, and magnetostrictive materials, the very low driv-
ing voltage of CNT is a major advantage for developing various actuator applications.
Other predicted future advantages are the direct conversion of electrical to mechanical
energy, high actuation strain, strength, and modulus.
The reference model of Fig. 15.13(a) can determine the strain in the actuator
due to the piezoresistive effect of the nanotube composite. We have demonstrated
this effect in which the electrical resistance of the material changes with strain:
Tension increases the resistance, and compression decreases the resistance. This
sensing idea can allow the same material to act as an actuator and as a strain sensor.
Actually there are three interesting effects acting together in the nanotube actuator:
the electrochemical actuation, the piezoresistive effect, and the electrochemical
FIGURE 15.13 Models of CNT actuators developed at University of Cincinnati: (a) self-
sensing CNT circuit; (b) controller design.
q t
t
x y
C E
V t
t
v
( )
( )
( )
( )
1/
Z
p R
d
R
d
Reference model
V
2
R
s
R
s
C
d
C
d
CNT actuator model
R
p
R
p
V
1
V
(b) (a)
Output
Input
Collocated output
feedback control
Structure with
CNT actuators
+

charge generation effect in which charge is generated by electrolyte flow over the
nanotube material without strain. These effects will be studied separately and then
a combined model will be discussed.
The density of the nanotubes in the SPE has an effect on the electrochemical
actuation. Well-aligned but separated carbon nanotubes produce a large surface area
supercapacitor with high power density. The capacitance of the CNT actuator SPE
material is C Q/V H
o
, H
r
(A
s
/t), where Q and V are the charge and voltage,
respectively, and H
o
, H
r
, A
s
, and t are the permittivity of free space, the relative
permittivity, the surface area of the electrode, and the separation of the plates,
respectively. In electrochemical double-layer capacitors (ECDLCs) the inter-plate
spacing t is reduced to the Helmholz double layer, which is 0.34 nm. The active
area of the actuator depends on the density of nanotubes in the composite and the
length and diameter of the nanotubes. A density of at least 50 million CNFs/cm
2
is
expected, and the density of SWCNTs will be much greater. The density achievable
will depend on how well the nanotubes can be functionalized and dispersed to
provide effective ion exchange and load transfer. The relative permittivity is depen-
dent on the type of nanotube including the chirality, number of walls, spacing
between the walls, and diameter of the nanotube. There have not been many studies
of the electrochemical properties of carbon nanotubes in a conductive polymer
composite. The transduction properties of the CNT material must be studied exper-
imentally to determine parameters to model the strain and to find the optimum
material and design to maximize the capacitance and the ion exchange.
Synthesis of arrays of parallel nanotubes is another approach to making nanotube
smart materials. Embedding planar arrays in layers will allow build-up of composite
materials, as shown in Fig. 15.14(a). Another actuator type can be made by attaching
FIGURE 15.14 Concepts for CNT actuators; (a) SPE hybrid actuator; (b) linear actuator
built with 1-mm bricks; and (c) tripod actuator 1 mm high where the rotation is exaggerated
for illustration. Background shows a 1-mm-high MWCNT array grown by Yi Tu of First
Nano Inc. The array is separated to show the nanotube density.
A/D converter
Load cell
Control
signal
Load cell
CNT
Solid
electrolyte
Electrodes
Laser
displacement
sensor
A/D converter
Building
brick
figure 1(c)
(b) Linear actuator (c) Tripod actuator (a) Composite actuator
Solid or liquid
electrolyte
CNT actuator
Electrode
building bricks of the array in series by bonding or direct growth in a nanofurnace,
as shown in Fig. 15.14(b). A single segment or building brick can be used to form a
multisegment tilting actuator. A concept for this actuator is shown in Fig. 15.14(c),
and the deflection is exaggerated. About 1% strain is expected and the angle of
deflection will depend on how closely the nanotube posts are spaced. The linear
or tilting segments may be connected to design active devices from the nanoscale
to the macroscale in size. Inch-size planar arrays can be synthesized and current
efforts are focused on increasing the length of MWCNTs. The tilting motion may
be beneficial for alignment and pointing applications. A computer control system
will be used to program the voltage for the actuation.
Based on these experiments, a concept for an electrochemical active catheter
has been devised, as shown in Fig. 15.15. Benjamin Franklin invented the first
catheter. If smart materials could be used to build the active catheter proposed and
other miniature manipulators, they would have many medical applications. The
catheter has potential for use in angioplasty, stent insertion, plaque removal,
hydrocephaly treatment, remote surgery, and other applications. Medical device
and surgical innovation is a growing area where nanoengineering can make con-
tributions. In particular, methods to increase actuation stroke are needed. One
concept is telescoping MWCNTs illustrated in Fig. 15.15.
15.7 FUTURE DIRECTIONS
FOR INTELLIGENT MATERIALS
This section presents ideas for (near) future applications of nanoscale intelligent
materials.
15.7.1 CARBON STRUCTURAL NEURAL SYSTEMS
Structural health monitoring (SHM) is a subarea of intelligent structures wherein
the integrity of a structure is monitored in real-time by an in situ sensor system.
Several methods of SHM using smart materials have been proposed in the literature
and include damage identification using the vibration response of a structure, fiber-
optic strain measurement, stress wave propagation techniques, and wireless MEMS.
FIGURE 15.15 Concept for an active catheter using CNT electrochemical actuation and self-
sensing using the principle of electrochemical actuation or telescoping actuation.
Link
Inner tube
CNT actuator
Outer tube
These techniques for in situ real-time SHM may not be cost-effective for applications
to large structures that have complex geometries and varying thicknesses and cur-
vatures such as joints, ribs, fasteners, hybrid materials, and highly damped devices
such as honeycomb sandwich structures. Although there are several types of SHM
systems being developed, practical on-board SHM systems should add the lowest
complexity with the highest benefit. This section describes a concept for a new
sensor system architecture
9
that can make practical the use of large numbers of
miniature sensors to monitor mechanical, civil, and environmental systems that can
benefit from highly distributed sensors and massively parallel processing using a
small number of channels of data acquisition. In biological systems, millions of
identical parallel nerves are used for sensing. This architecture allows billions of
bits of sensory information to be processed in the neural, auditory, and visual systems
in an efficient hierarchical order and in harmony and researchers are attempting to
replicate the mechanisms of this biological system. This intersection of biological
function with nanoscale perfection may lead to an efficient and intelligent structural
neural system (SNS).
202
One of many possible architectures for the SNS is in the form of an array of
orthogonal intersecting nerves built with carbon nanotubes. This approach is
based on a design in which piezoceramic receptors are used to detect acoustic
emissions or high strains caused by damage propagation.
9,20,2527
The carbon
nanotube SNS works broadly on the principle of the biological nervous system.
27
The sensor fibers act as neurons and each neuron is connected to other neurons.
In the event of high strain or damage to the material, the piezoresistive sen-
sor/neuron changes resistance and changes the voltage of the neuron. In a bio-
logical system, the signal from one neuron is transmitted by special molecules
called neurotransmitters that cross the synaptic gaps to receptor sites on the target
neurons. The signal from a biological neuron is inhibitory or excitatory, not both.
Excitatory signals tend to fire the target neurons, and inhibitory signals tend to
prevent firing. In our SNS, the firing of the neuron depends on the threshold
voltage of the sensor. If the voltage is above a certain level, the sensor/neuron
sends an excitatory signal that in turn combines with other firing neurons, and
thus the signal is transmitted. Additional layers of hierarchy are possible. If the
sensor/neuron is far away from the damage event, it will generate a voltage that
is below the threshold, so the neuron will not be allowed to fire and it will act
as an inhibitory signal to subsequent neurons. The threshold can be determined
by having a capacitor/resistor in the circuit. This is analogous to synaptic trans-
mission of a signal in a biological nervous system. The signal from the closest
sensor/neuron can be processed to extract the information about the damage. The
other neurons can be prevented from firing (occluded) to eliminate unwanted
signals. The threshold voltage can be determined by experimentation and simu-
lation, depending on the size and type of SNS. This approach uses receptors that
cross but electrically independent. This allows individual row and column neurons
to be formed in horizontal rows and vertical columns, as shown in Fig. 15.16. A
single neuron is shown in Fig.15.4(a).
A unit cell of the SNS consists of two row receptors and two column receptors.
When damage occurs near one receptor, the row and columns of that receptor fire
and the other rows and columns are inhibited or occluded, depending on the design
and training of the neurons. The intersection of the rows and columns that are firing
will show the location of the damage. Also any change in the resistance of the
neuron can be related to the strain and damage in the structure. This system is
redundant and minimizes false signals. If damage causes one neuron to become
inoperable, a problem would be indicated by the failed neuron, and the damage
would also be detected when it propagates to the neighboring neuron. Since the
neurons can be scaled to the micron or smaller level, an arbitrary degree of sensi-
tivity can be obtained.
This SNS can use low frequency data acquisition, but requires a very high sensor
density to detect initiating damage. For example, crack propagation around rivets
requires a small-scale unit cell around each rivet. This is practical due to the small
size and high strain sensitivity of CNT strands. No diagnostic signal or amplifiers or
large sets of historical data are needed in this simple approach. In addition, the size
of the carbon SNS can be miniaturized such that the nerves can be put inside
composite materials or at the bonding interfaces of joints without affecting the integ-
rity of the structural material. This approach overcomes many of the problems of
embedding sensors inside composites. The piezoresistive sensors are highly distrib-
uted, simple, less expensive, less intrusive, and more sensitive than existing SHM
sensor technology. Figure 15.16(a) shows biological neurons, and Fig. 15.16(b)
shows an array SNS with ten row and ten column neurons (N1N20), which are
continuous sensors that sense along their entire length. The row and column neurons
in the array are routed into four signal channels two channels of time response
signals and two channels for neuron firing. Using information from these four
channels, it is possible to locate the damage source and measure the signal amplitude.
The analog to digital conversion is only needed at the PC level, and this reduces the
complexity of the instrumentation and the cost, size, and weight of the SHM system.
It is anticipated that continued development of the array SNS using nanotube-based
sensors will lead to ultra high-density high sensitivity sensor systems that can be
FIGURE 15.16 The biological neural system and an array carbon structural neural system.
(c) Intelligent aircraft (a) Biological neurons (b) Carbon nanotube array neurons
N1 N10
N11
N20
Channels 1,2
Columns output
(Neurons 1-10)
Channels 3,4
Rows output
(Neurons 11-20)
Processors
Square unit cell
Panel
structure
Continu-
ous
sensor
Data acquisition system
and signal processing
On board computer
for damage detection
and warning
Brain
Human
nervous
system
Artificial nervous
system
Human body
Advanced aircraft
built at either the macroscopic or the microscopic scale for use on safety-critical
structures such as aircraft (Fig. 15.16(c)). The SNS architecture may also have
applications in health monitoring of living systems and the environment.
15.7.2 HIGH-TEMPERATURE NANOSCALE MATERIALS
Since piezoelectric ceramic materials have been under development for a long time,
new materials are sought that can provide increased performance at elevated tem-
peratures. There are several nanoscale materials under development that may be
suitable for sensing at high temperatures. These materials are in early stages of
development and are briefly discussed next. References are given to allow readers
to obtain more information.
Nanocomposites with a0.5 to 1 wt% of SWCNTs can become self-sensing
structures by the addition of electrodes to the composite surfaces, but a host material
that can withstand high temperatures is needed. The piezoelectric effect is small in
SWCNTs, and it does not seem practical for use in sensing. The electrochemical
effect of the SWCNT is based on double-layer charge injection that causes bond
expansion. A limitation is that the nanotubes must have an ion exchange for the
double-wall charge injection to work. For high temperature applications, a molten
electrolyte might be used as suggested by Baughman et al.
53
The SWCNTs are
formed at temperatures from 700 to 1000qC. They are expected to operate at tem-
peratures up to 500qC, beyond which they can burn in an oxygen environment. Use
of a molten electrolyte would pose problems in some applications.
Noncarbon nanotubes have received much less attention in recent years and
remain relatively unexplored. The BNTs are predicted to be piezoelectric in bundles
and the maximum piezoelectric-induced stress constant occurs for the specific (5,0)
chirality, but the piezoelectric d coefficient is about 20 times smaller than for piezo-
ceramics. (See Chapter 21.) BNT materials are the subjects of different investigations
that started around 2000 and their characteristics offer the possibility of piezoelectric
actuation tailored over a wide bandwidth, and the possibility of replacing piezocer-
amic materials with BNTs. However, the synthesis of BNTs is at an early stage (see
Chapter 21) and the information on this issue in the scientific literature is limited.
There appears to be no experimental verification of the properties of the BNTs, and
it is not possible to control the chirality of the nanotubes produced. BNTs are formed
at temperatures above 1000
o
C. They are expected to operate at temperatures up to
1000qC. BNTs represent one possible material for high temperature applications.
Silicon carbide nanotubes (SiCNTs) are under early stages of synthesis at NASA
Glenn, as described on the web site.
10
Macroscopic SiC is a very hard and heat-
resistant ceramic material. The objective of this approach is to take advantage of
the properties of SiC to synthesize and characterize SiCNTs for high temperature
and high radiation conditions. Multiple synthetic approaches are being considered
that parallel the direct CNT formation, as well as an indirect approach involving
derivatization of a CNT to a SiCNT. Recently SiCNTs were produced in-house at
NASA Glenn. A CVD-template synthesis method, in which the template functions as
a nanomold, was employed to produce highly ordered arrays of SiCNTs. The arrays
are composed of individual SiCNTs (200 nm in diameter, 60 Pm in length) aligned
vertically, with a density of approximately 109 tubes/cm
2
. The electrical and mechan-
ical properties of the SiCNTs are being characterized and compared with theoretical
SiCNT modeling results. Electrical properties studies of nanotubes at high temper-
atures could determine the suitability of nanotube high-temperature sensors. The
electrical activity of SiCNTs could also be studied as a function of adsorbates, which
could ultimately lead to applications such as nanogas sensors for harsh environments
and sensors for the working areas of jet engines.
Bismuth nanowires (see Chapter 3), ZnO nanobelts (see Chapter 4) and other
inorganic materials such as WS
2
MoS
2
, NbS
2
, and TiS
2
may have piezoelectric or
other interesting properties that might be useful as sensors.
In summary, existing piezoelectric materials do not have satisfactory properties
at high temperatures. Some of the different forms of nanotubes that are naturally polar
in bundles or otherwise fullerene-like may be useful for high temperature applica-
tions. Further research is needed to develop and test the piezoelectric nanotubes.
15.7.3 POWER HARVESTING USING CARBON NANOTUBES
In preliminary experiments we tested an induced charge generator in an electrolyte
for power harvesting in mechanical and structural systems. The CNT material is
shown in Fig. 15.17(a). The voltage due to vertical motion of the film in the
electrolyte was a 7-mV peak in the electrolyte. The film was taken out of the
electrolyte and placed in deionized water and the initial peak voltage was 17 mV, as
shown in Fig. 15.17(b), but upon continued cycling, the voltage reduced to near
zero. This result showed that a micron thin film with a volume of 5 mm by 20 mm
by 0.018 mm = 1.8 mm
3
can generate a small amount of power (a10 nW) due to
electrolyte flow over the nanotubes, not by strain of the nanotubes. It was also
determined that the flow direction should be in the long direction of the nanotube
actuator to generate the greatest voltage. A bending type motion produced a 2-mV
signal. These are interesting results that are partly discussed in previous litera-
ture.
157158
This experiment showed that charge may be partially shunted due to the
ionic conductivity of the electrolyte, and that increased power generation might be
obtained by partially encapsulating a hydrated material to prevent charge loss
through the electrolyte. This is currently under investigation; see Reference 66. It
is also interesting that the charge produced is flow dependent and greater when the
flow is in the long direction of the nanotube paper. The flow in the transverse
direction yields only 12 mV. The width of the film, electrode location, and some
alignment of the nanotubes in the long direction may explain this result. One
possible application of the nanotube film is to have a self-powered biosensor made
of the patterned nanotube array described earlier. This self-contained sensor could
be inside the body to detect diseases and monitor critical signals from the body
and transmit signals wirelessly from inside the body using power harvested due to
blood flow through a CNT array. The MWCNT array might also simultaneously
act as a cancer sensor and generate power to send signals wirelessly. Figure 15.17(c)
shows an approach planned to demonstrate the power generation of the smart
material. A flow-induced charge generator MWCNT array used for power harvest-
ing in ionic fluidic systems is shown. Flow or horizontal vibration will be applied
to the CNT material, and a circuit will be used to measure the power developed.
An RF module may be used to test wireless transmission for later use in a SHM
system. The CNT power generator can power a batteryless RF module that transmits
a short-range wireless ID code to a nearby server. Using the RF module, structural
states, vibration, strain, or corrosion can be monitored wirelessly and in real time.
While monitoring the structural state, we can identify sensors sending an abnormal
signal, since every sensor has its own identification. A LABVIEW system will be
used to control the power generation experiment.
15.7.4 INTELLIGENT MACHINES
The intelligent machine is a new concept for manufacturing, self-repair, and demanu-
facturing based on the use of nanoscale materials. One concept is to develop a carbon
nanocomposite material for assembly and disassembly of machines and to fabricate
parts used to manufacture machines. The nanocomposite material is produced by
incorporating carbon nanotubes and carbon nanofibers into a polymer matrix. The
FIGURE 15.17 Power generation by vertical fluid flow over a SWCNT Nafion film: (a) film
on two sides of a stiff backing; (b) initial 17 mV generated in deionized water after 7 mV
generated in 5 M NaCl; and (c) concept for flow-induced power harvesting with six stacked
MWCNT arrays.
(b)
(a)
Ionic
fluid
flow
Regulator Storage
Porous
substrate
Full wave
rectifier
MWCNT Array
PIC
Control ler
RF
Module
MWCNT
array
PIC
controller
RF
module
(c)
nanocomposite material has piezoresistive sensing to measure large strains and detect
damage in parts. The material also has electrical conductivity that can be used to
heat and recure the material in the appropriate area to partially self-repair damage,
and also to heat the material to disassemble a product when it reaches the end of
its useful life, either due to wear or obsolescence. A machine with the capabilities
for sensing damage, partial self-repair of damage, and controlled disassembly is
described here as an intelligent machine.
Carbon nanotubes added to polymers can be viewed as new smart materials
that have satisfactory strength, sensing, and actuation capabilities. A novel appli-
cation of this nanocomposite smart material is to form machine parts or for use as
an adhesive to assemble parts. This nanocomposite material can be turned on to
repair parts by electrical heating to recure and heal the material in the damaged
area, or by heating to debond and partially disassemble the product to replace or
repair parts. Also, at the end of the product life, the smart material can be heated
to debond and allow complete disassembly of the product efficiently and without
damaging reusable or hazardous components. The demanufactured components will
be sorted and either recycled or remanufactured into new products. Carbon nanotube
polymers have recently been used at UC to form different structural smart materials
with electrical conductivity, sensing, and actuation capabilities. The properties of
these materials can be characterized and optimized for applications related to
intelligent machines.
The benefit of utilizing nanotechnology in manufacturing is that manufacturing
and consumer use of products would change from traditional use of fasteners for
assembly to quick assembly using adhesives. Repair of damage would be done by
reforming and curing the material, and large structural parts could be made of a
nanocomposite material that is self-sensing and self-repairing. In addition, discarding
old products would cease and be replaced by a new paradigm in which the purchaser
would receive a credit to return obsolete or worn-out products to the manufacturer.
A reverse manufacturing process, defined here as demanufacturing, would use a
disassembly line to apply the appropriate stimulation to the part to disassemble itself
into useful smaller components that would subsequently be sorted and reused,
disposed of, or shipped to other remanufacturing facilities. If this new manufacturing
ideology is successfully developed, it will have a large impact on manufacturing in
the following areas: (1) reducing the costs of manufacturing by efficiently reusing
parts or materials from old products; (2) complex machines could be designed to
be assembled faster with less need for critical tolerances; (3) parts count could be
reduced using adhesive joining; (4) components could be disassembled easily for
repair by electrical heating in some cases combined with ultrasound, which would
eliminate fasteners for many applications; (5) machines assembled using smart
materials and parts from old machines that could be efficiently disassembled using
smart materials could be remanufactured at a cost lower than the cost of a new
machine; (6) demanufacturing would promote conservation by reusing materials and
reduce overall product costs; and (7) the new paradigm of intelligent machines
could revitalize manufacturing industries because new types of products would be
produced and recycled. In addition, manufacturing plants are located almost everywhere
and the new idealogy would provide a good opportunity to educate students and
bring nanotechnology to industry. As an example, students in cooperative education
programs could work in industries and government laboratories to investigate the
use of smart materials for manufacturing and demanufacturing of different products.
15.7.5 TELESCOPING CARBON NANOTUBES
Carbon nanotube polymer materials being developed have electrical and ionic con-
ductivity properties wherein the polymer is actually a semiconducting structural mate-
rial that can function as a sensor and actuator. There is another concept for a potential
actuation and sensing mechanism that can be used with a conductive polymer elec-
trolyte material. This concept mechanism is telescoping of MWCNTs. Telescoping of
the inner tubes in a MWCNT that has the outer tube opened has been shown using
an atomic force microscope.
212
It has also been suggested that telescoping and oscil-
lating actuation is possible and the MWCNTs have been proposed for use as rotational
and translational bearings, and as nuts and screws for building nanomachines by taking
advantage of the spiral chirality of nanotubes.
209-217
Because of the potentially huge
displacement of this type of actuator, telescoping nanotube arrays (TNAs) are of
interest for linear actuation applications based on electrostatic repulsion actuation. The
telescoping nanotube has also been shown to change electrical resistance with tele-
scoping length.
209, 211
Thus, the TNA can also become a displacement sensor. Moreover,
based on our testing of MWCNT power cells to generate electricity, we believe the
TNA can also be used for power generation due to telescoping of the material when
it is encapsulated in an electrolyte. The electrically telescoping nanotube has not been
demonstrated and is an area of current investigation and future work at the University
of Cincinnati and elsewhere.
15.8 CONCLUSIONS
The attempt in this chapter was to provide an up-to-date review of nanotube intelligent
materials, structures, and applications. Special nanotube properties were explained,
including electrical conductivity, contact resistance, piezoelectricity, piezoresistance,
and electrochemical actuation by charge injection that might be useful to generate
ideas for new intelligent materials and structures. Various types of nanostructured smart
sensors and actuators were also proposed. Although significant progress has been made
in understanding nanoscale smart materials, there is much still to be learned, especially
about bringing the nanoscale properties to the macroscale. It can, however, be con-
cluded that nanoscale intelligent materials will engender fundamental changes in the
field of smart materials and structures in the near future, and there are some emarkable
possibilities for building intelligent structures using nanotechnology.
PROBLEMS
1. In the chapters in this book, different types of nanoscale materials are
discussed. These include carbon nanotubes, boron nitride nanotubes, zinc
oxide nanobelts, and bismuth nanowires. What might be the advantages
and disadvantages of building smart materials using these different types
of nanoscale materials?
2. Describe the principle of the Potentiostat/Galvanostat. Describe the satu-
rated Calomel reference electrode (SCE). You may go to the manufactures
Websites for information on electrochemistry.
3. What is a motor protein? Why are DNA and motor proteins possibly good
actuators in vivo? Can a biological smart machine be made?
4. Define solid polymer electrolyte. What is the limitation of a solid polymer
electrolyte in a structural application? Propose ways to improve the
mechanical properties of solid polymer electrolytes.
5. Describe the sp
2
hybridization in the CNT structure and the V-bonds and
S-bonds.
6. A machine with the ability to prevent failure is a smart machine. A
manufacturer wants to develop a way to provide short-time emergency
lubrication for the bearings in an aircraft engine. This would allow the
aircraft to operate long enough to safely land if the normal bearing lubri-
cation system fails. Think of a nanoscale material that could be a lubricant.
Design a system that automatically delivers this lubricant to the bearing
for a short time if the standard lubrication system fails.
7. Search the literature for the properties of nanoscale materials and try to
compute and fill in the missing properties in Table 15.3.
ACKNOWLEDGMENTS
This work was sponsored in part by the University of Cincinnati (UC) Summer
Student Fellowship, the universitys Research Council, FirstNano, Inc., the
National Renewable Energy Laboratory, and the Ohio Aerospace Institute. This
funding support is gratefully acknowledged. In addition, the following student
researchers assisted in the development of nanoscale and smart materials at UC:
Courtney Brown, Nicole Reinart, Jacob Hause, Jennifer Chase, Payal Kaul, Robert
Gilliland, and Subrahmin Srivinas. Other help was provided by Chris Sloan, Dennis
Adderton, and Yi Tu of FirstNano, and Bo Westheider, Dale Weber, George
Kreishman, Larry Schantman and Dave Breiheim of UC. Assistance in the catheter
concept was provided by doctors Randall Wolf and Charles Dorn of the Department
of Surgery, Center for Surgical Innovation at UC. Drs. Zhongyun Dong and Abdul
Rahman Jazieh of the Department of Internal Medicine at UC provided the ideas
for using antibodies for cancer detection. Hong-Chao Zhang of the Center for
Applied Research in Advanced Manufacturing, Lubbock, TX, Jay Lee of the
University of Wisconsin, and John Sutherland provided ideas for the smart
machine. All their help and cooperative work are gratefully acknowledged and
made this chapter possible.
REFERENCES
1. Carbon Nanotechnologies, Inc., Houston, TX, www.cnanotech.com; contact person,
Tom Pittstick.
2. NanoLab, Inc., info@nano-lab.com.
3. Applied Sciences, Inc. and Pyrograf Products, Inc.; contact person, David Burton.
4. Dresselhaus, M.S., Dressehaus, G., and Eklund, P.C., Science of Fullerenes and
Carbon Nanotubes, Academic Press, San Diego, 1996.
5. FirstNano, Inc., Santa Barbara, California, 93111.
6. Dresselhaus, M.S., Dresselhaus, G., and Avouris, Ph., Carbon nanotubes: synthesis,
structure, properties, and applications, in Springer, 80th Topics in Applied Physics,
2000, Springer-Verlag, New York.
7. Goddard, W.A., Brenner, D.W., Lyshevski, S.E., and Lafrate, G.J., Handbook of
Nanoscience, Engineering, and Technology, CRC Press, Boca Raton, FL, 2003.
8. Nalwa, H.S., Nanostructured Materials and Nanotechnology, Academic Press, San
Diego, 2000.
9. Smart Structures Bio-Nanotechnology Laboratory, University of Cincinnati, OH,
http://www.min.uc.edu/~mschulz/smartlab/smartlab.html.
10. NASA Glenn Research Center. http://www.grc.nasa.gov/WWW/SiC/redhot.html
11. Department of Electronic Materials Engineering, The Australian National Univer-
sity, Canberra. http://www.anutech.com.au/TD/Info%20Sheets/ Nanotubes
%20brochure.pdf
12. Mallick, P.K., Fiber Reinforced Composites, Marcel Dekker, New York, 1988.
http://www.hexcelfibers.com/Markets/Products/default.htm
13. Cady, W.G., Piezoelectricity: An Introduction to the Theory and Applications of
Electromechanical Phenomenon in Crystals, Dover, New York, 1964.
14. Jaffe, B. and Cook, W.R., Jr., Piezoelectric Ceramics, Academic Press, San Diego,
1971.
15. Uchino, K., Piezoelectric Actuators and Ultrasonic Motors, Kluwer Academic Pub-
lishers, Boston, 1966.
16. APC International, Piezoelectric Ceramics: Principles and Applications, http://
www.americanpiezo.com/materials/index.html
17. Dubois, M.-A. and Muralt, P., Measurement of the effective transverse piezoelectric
coefficient e
31,f
of AIN and Pb(Zr
x
,Ti
1-x
)O
3
thin films, Sensors Actuators A, 77,
106112, 1999.
18. Eitel, R.E., Randall, C.A., Shrout, T.R., Rehrig, P.W., Hackenberger, W., and Park,
S., New high temperature morphotropic phase boundary piezoelectrics based on
Bi(Me)O
3
-PbTiO
3
ceramics, Jpn. J. Appl. Phys., 40, 59996002, 2001.
19. Schulz, M.J., Sundaresan, M., McMichael, J., Clayton, D., Sadler, R., and Nagel, W.,
2003, Piezoelectric materials at elevated temperature, J. Intelligent Mater. Syst.
Struct., 14, 11, 693705.
20. Martin, W.N., Ghoshal, A., Schulz, M.J., and Sundaresan, M.J., Structural health
monitoring using an artificial neural system, Recent Res. Dev. Sound Vibration, 2003.
21. Bent, A.A., Active Fiber Composites for Structural Actuation, Ph.D. dissertation,
Massachusetts Institute of Technology, Cambridge, MA, 1997.
22. Bent, A.A and Hagood, N.W., Piezoelectric fiber composites with interdigitated
electrodes, J. Intell. Mater. Syst. Struct., 8, 1998.
23. CeraNova Corporation, Franklin, MA.
24. Wilkie, K., NASA LaRC Macro-Fiber Composite Actuator, NASA News Release,
September 19, 2000.
25. Datta, S., An Active Fiber Continuous Sensor for Structural Health Monitoring, M.S.
thesis, University of Cincinnati, 2003.
26. Sundaresan, M.J., Ghoshal, A., and Schulz, M.J., Sensor Array System, U.S. patent
6,399,939 B1, 2002.
27. Sundaresan, M.J., Schulz, M.J., Ghoshal, A., and Pratap, P., A Neural System for
Structural Health Monitoring, paper presented at SPIE 8th Symposium on Smart
Materials and Structures, March, 48, 2001.
28. Piezo-Systems, Inc. www.piezo.com, and, http://www.piezo.com/psi5a4.html.
29. Measurement Specialties. http://www.msiusa.com/default/index.asp.
31. EasyTube CVD Nanofurnace, FirstNano, Inc., Santa Barbara, CA. http://www.first-
nano.com.
32. Hafner, J.H., Bronikowski, M.J., Azamian, B.R., Nikolaev, P., Rinzler, A., Colbert,
D.T., Smith, K.A., and Smalley, R.E., Catalytic growth of single-wall carbon nano-
tubes from metal particles, Chem. Phys. Lett., 296, 195202, 1998.
33. Huang, S., Cai, X., and Liu, J., Growth of millimeter long and horizontally aligned
single-walled carbon nanotubes on flat substrates, JACS, 125, 56365637, 2003.
34. Wei, B.Q., Vajtai, R., Zhang, Z.J., Ramanath, G., and Ajayan, P.M., Carbon nanotube-
magnesium oxide cube networks, J. Nanosci. Nanotechnol., 1, 3538, 2001.
35. Buldum, A. and Lu, J.P., Contact resistance between carbon nanotubes, Phys. Rev.
B, 63, 161403, 2001.
36. Tian, W. and Datta, S., Aharonov-Bohm-type effect in graphene tubules: a Landauer
approach, Phys. Rev. B, 49, 50975100, 1994.
37. Ananntram, M.P., Datta, S., and Xue, Y., Coupling of carbon nanotubes to metallic
contact, Phys. Rev. B, 61, 20, 14219, 1999.
38. Mortensen, N.A., Johnsen, K., Jauho, A.-P., and Flensberg, K., Contact resistance of
quantum tubes, Superlattices Microstructures, 26, 6, 351361, 1999.
39. Bachtold, A., Henny, M., Terrier, C., Strunk, C., Schonenberger, C., Salvetat, J.P.,
Bonard, J.M., and Forro, L., Contacting carbon nanotubes selectively with low ohmic
contacts for four-probe electric measurements, Appl. Phys. Lett., 73, 2, 274276,
1998.
40. Menon, M., Andriotis, A., and Froudakis, G., Curvature dependence of metal catalyst
atom interaction with carbon nanotubes wall, Chem. Phys. Lett., 320, 42434, 2000.
41. Tachibana, T., Williams, B., and Glass, J., Correlation of the electrical properties of
metal contacts on diamond films with the chemical nature of the metal-diamond
surface 2, titanium contacts: a carbide forming metal, Phys. Rev. B, 45, 1197511981,
1992.
42. Zhang, Y., Franklin, N., Chen, R., and Dai, H., A metal coating study of suspended
carbon nanotubes and its implications to metal-tube interactions, Chem. Phys. Lett.,
331, 3541, 2000.
43. Benoit, J.M., Buisson, J.P., Chauvet, O., Godon, C., and Lefrant, S., Low-frequency
Raman studies of multiwalled carbon nanotubes: experiments and theory, Phys. Rev.
B, 66, 073417, 2002.
44. A. Ahuwalia, R., Baughman, D., De Rossi, A., Mazzoldi, M., Tesconi, A., Tongnetti,
G., Vozzi, Microfabricated electroactive carbon nanotube actuators, Proc. SPIE Conf.
Smart Struct. Mater., 4329, 2001.
45. Jain, S., Kang, P., Yeo-Heung, Y., He, T., Pammi, S.L., Muskin, A., Narsimhadevara,
S., Hurd, D., Schulz, M.J., Chase, J., Subramaniam, S., Shanov, V., Boerio, F.J., Shi,
D., Gilliland, R., Mast, D., and Sloan, C., Developing Smart Materials Based on
Carbon Nanotubes, paper presented at SPIE 11th International Symposium on Smart
Structures, San Diego, 2004.
46. Gao, M., Dai, L., Baughman, R.H., Spinks, G.M., et al., Electrochemical properties
of aligned nanotube arrays: basis of new electromechanical actuators, electroactive
polymer actuators and devices, SPIE Proceedings, 1824, 2000.
47. Mazzoldi, D.D. and Baughman, R.H., Electromechanical behavior of carbon nanotube
sheets in electrochemical actuators: electroactive polymer actuators and devices, SPIE
Proceedings, 2532, 2000.
48. El-hami, K. and Matsushige, K., Covering single walled carbon nanotubes by the
PVDF copolymer, Chem. Phys. Lett., 368, 168171, 2003.
49. Knez, M., Sumser, M., Bittner, A.M., Wege, C., Jeske, H., Kooi, S., Burghard, B.,
and Kern, K., Electrochemical modification of individual nano-objects, J. Elec-
troanaly Chem., 522, 7074, 2002.
50. Baughman, R.H., Conducting polymer artificial muscles, Synth. Met., 78, 339353,
1996.
51. Dickinson, M.H., Farley, C.T., Full, R.J., Koehl, M.A.R., Kram, R., and Lehman, S.,
How animals move: an integrative view, Science, 288, 100106, 2003.
52. Inganas, O. and Lundstrum, I., Carbon nanotube muscles, Science, 284, 3132, 1999.
53. Baughman, R.H., Cui, C., Zakhidov, A.A., Iqbal, Z., Barisci, J.N., Spinks, G.M.,
Wallace, G.G., Mazzoldi, A., De Rossi, D., Rinzler, A.G., Jaschinski, O., Roth, S.,
and Kertesz, M., Carbon nanotube actuators, Science, 284, 13401344, 1999.
54. Fraysse, J., Minett, A.I., Gu, G., Roth, S., Rinzler, A.G., and Baughman, R.H.,
Towards the demonstration of actuator properties of a single carbon nanotube, Curr.
Appl. Phys., 1, 407411, 2001.
55. Minett, A., Fraysse, J., Gang, G., Kim, G.-T., and Roth, S., Nanotube actuators for
nanomechanics, Curr. Appl. Phys., 2, 6164, 2002.
56. Fraysse, J., Minett, A.I., Jaschinski, O., Dueshberg, G.S., and Roth, S., Carbon
nanotubes acting like actuators, Carbon, 40, 17351739, 2002.
57. Spinks, G.M., Wallace, G.G., Lewis, T.W., Fifield, L.S., Dai, L., and Baughman,
R.H., Electrochemically driven actuators from conducting polymers, hydrogels, and
carbon nanotubes, Proc. SPIE Int. Soc. Opt. Eng. 4234 (Smart Materials), 223231,
2001.
58. Barisci, J.N., Spinks, G.M., Wallace, G.G., Madden, J.D., and Baughman, R.H.,
Increased actuation rate of electromechanical carbon nanotube actuators using poten-
tial pulse with resistance compensation, Smart Mater. Struct., 12, 549555, 2003.
59. Baughman, R.H., Conducting polymer artificial muscles, Synth. Met., 78, 339353,
1996.
60. Tahhan, M., Truong, V.-T., Spinks, G.M., and Wallace, G.G., Carbon nanotube and
polyaniline composite actuators, Smart Mater. Struct., 12, 626632, 2003.
61. http://www.mpi-stuttgart.mpg.de/klitzing/research/nano/nanoactuators.html
62. http://www.nasatech.com/Briefs/Sep01/NPO21153.html
63. http://www.cs.unc.edu/Research/nano/tams.html
64. Roth, S. and Baughman, R.H., Actuators of individual carbon nanotubes, Curr. Appl.
Phys., 2, 311314, 2002.
65. Livage, J., Actuator materials: towards smart artificial muscle, Nature Mater., 2,
297299, 2003.
66. Inpil Kang, Yun Yeo Heung, Jay H. Kim, Jong Won Lee, Ramanand Gollapudi,
Srinivas Subramaniam, Suhasini Narasimhadevara, Douglas Hurd, Goutham R.
Kirikera, Vesselin Shanov, Mark J. Schulz, Donglu Shi, Jim Boerio, Carbon nanotube
smart materials, submitted, Comp. B J.
67. Yun Yeo-Heung, Atul Miskin, Phil Kang, Sachin Jain, Suhasini Narasimhadevara,
Douglas Hurd, Mark J. Schulz; Vesselin Shanov, Tony He, F. James Boerio, Donglu
Shi, Subrahmin Srivinas; Carbon nanofiber hybrid actuators. Part I: Liquid electrolyte-
Based, J. Int. Mater. Smart Struct., submitted, special issue.
68. Yun Yeo-Heung, Atul Miskin, Phil Kang, Sachin Jain, Suhasini Narasimhadevara,
Douglas Hurd, Mark J. Schulz, Vesselin Shanov, Tony He, F. James Boerio, Donglu
Shi, Subrahmin Srivinas; Carbon nanofiber hybrid actuators. Part II: solid electrolyte-
based, J. Int. Mater. Smart Struct., submitted, special issue.
69. Davis, V.A., Ericson, L.M., Saini, R., Sivarajan, R., Hauge, R.H., Smalley, R.E., and
Pasquali, M., Rheology, phase behavior, and fiber spinning of carbon nanotube dis-
persions, AIChe Annual Meeting, 2001.
70. Girifalco, L.A., Hodak, M., and Lee, R.S., Carbon nanotubes, buckyballs, ropes, and
a universal graphitic potential, Phys. Rev. B, 62, 1310413110, 2000.
71. Liu, H.T., Cheng, F.L., Cong, G.S., et al., Synthesis of macroscopically long ropes
of well-aligned single-walled carbon nanotubes, Advan. Mater., 12, 11901192, 2000.
72. Walters, M.J., Casavant, X.C., Qin, C.B.H., Boul, P.J., et al., In-plane-aligned mem-
branes of carbon nanotubes, Chem. Phys. Lett., 338, 1420, 2001.
73. Shenton, W., Douglas, T., Young, M., Stubbs, G., and Mann, S., Inorganic-organic
nanotube composites from template mineralization of tobacco mosaic virus, Advan.
Mater., 11, 3, 253256, 1999.
74. Li, D., Zhang, X., Sui, G., Wu, D., and Liang, J., Toughness improvement of epoxy
by incorporating carbon nanotubes in the resin, J. Mater. Sci. Lett., 22, 791793, 2003.
75. Rege, K., Raravikar, N.R., Kim, D.-Y., Schadler, L.S., Ajayan, P.M., and Dordick,
J.S., Enzyme-polymer-single walled carbon nanotube composites as biocatalytic
films, Nano Lett., 3, 6, 829832, 2003.
76. Schaefer, D.W., Zhao, J., Brown, J.M., Anderson, D.P., and Tomlin, D.W., Morphol-
ogy of dispersed carbon single-walled nanotubes, Chem. Phys. Lett., 375, 369375,
2003.
77. Watts, P.C.P., Hsu, W.K., Kotzeva, V., and Chen, G.Z., Fe-filled carbon nanotube-
polystyrene:RCL composites, Chem. Phys. Lett., 366, 4250, 2002.
78. Dujardin, E., Ebbesen, T.W., Krishnan, A., and Treacy, M.M.J., Wetting of single
shell carbon nanotubes, Advan. Mater., 10, 17, 14721475, 1998.
79. Lozano, K., Bonilla-Rios, J., and Barrera, E.V., A study on nanofiber-reinforced
thermoplastic composites (2): Investigation of the mixing rheology and conduction
properties, J. Appl. Polym. Sci., 80, 11621172, 2001.
80. Hughes, M., Shaffer, M.S.P., Renouf, A.C., Singh, C., Chen, G.Z., Fray, D.J., and Windle,
A.H., Electrochemical capacitance of nanocomposite films formed by coating aligned
arrays of carbon nanotubes with polypyrrole, Advan. Mater., 14, 5, 382385, 2002.
81. Chen, G.Z., Shaffer, M.S.P., Coleby, D., Dixon, G., Zhou, W., Fray, D.J., and Windle,
A.H., Carbon nanotube and polypyrrole composites: coating and doping, Advan.
Mater., 12, 7, 522526, 2000.
82. Ounaies, Z., Park, C., Wise, K.E., Siochi, E.J., and Harrison, J.S., Electrical properties
of single wall carbon nanotube reinforced polyimide composites, Composites Sci.
Technol., 63, 2003.
83. Vander Wal, R.L. and Hall, L.J., Nanotube coated metals: New reinforcement mate-
rials for polymer matrix composites, Advan. Mater., 14, 18, 10341308, 2002.
84. Tahhan, M., Truong, V.-T., Spinks, G.M., and Wallace, G.G., Carbon nanotube and
polyaniline composite actuators, Smart Mater. Struct., 12, 626632, 2003.
85. Stephan, C., Nguyen, T.P., Lamy de la Chapelle, M., Lefrant, S., Journet, C., and
Bernier, P., Characterization of single walled carbon nanotubesPMMA composites,
Synth. Met., 108, 139149, 2003.
86. Fan, J., Wan, M., Zhu, D., Chang, B., Pan, W., and Xie, S., Synthesis and properties
of carbon nanotube-polypyrrole composites, Synth. Met., 102, 12661267, 1999.
87. Byron Piper, R., Frankland, S.J.V., Hubert, P., and Saether, C., Self-consistent prop-
erties of carbon nanotubes and hexagonal arrays as composite reinforcements, Com-
posites Sci. Technol., 63, 13491358, 2003.
88. Shinoda, H., Oh, S.J., Geng, H.Z., Walker, R.J., Zhang, Z.B., McNeil, L.E., and Zhou,
O., Self-assembly of carbon nanotubes, Advan. Mater., 14, 12, 899901, 2002.
89. Chen, J.H., Huang, Z.P., Wang, D.Z., Yang, S.X., Wen, J.G., and Ren, Z.F., Electro-
chemical synthesis of polypyrrole/carbon nanotube nanoscale composites using well-
aligned carbon nanotube array, Appl. Phys. A, 73, 129131, 2001.
90. Islam, M.F., Rojas, E., Bergey, D.M., Johnson, A.T., and Yodh, A.G., High weight
fraction surfactant solubilization of single-wall carbon nanotubes in water, Nano Lett.,
3, 2, 2003.
91. Nardelli, M.B., Polarization Effects in Nanotube Structures, presentation notes, North
Carolina State University. http://nemo.physics.ncsu.edu/~nardelli/index.html
92. Lebedev, N.G., Zaporotskova, I.V., and Chernozatonskii, L.A., On the Estimation of
Piezoelectric Modules of Carbon and Boron Nitride Nanotubes, Volograd State Uni-
versity and Institute of Biochemical Physics, Moscow, Russia, 2001.
93. Rubio, A., Corkill, J., and Cohen, M., Theory of graphitic boron nitride nanotubes,
Phys. Rev. B, 49, 5081, 1994.
94. Bando, Y., Ogawa, K., and Golberg, D., Insulating nanocables: Invar Fe-Ni alloy
nanorods inside BN nanotubes, Chem. Phys. Lett., 347, 349, 2001.
95. Erkoc, S., Structural and electronic properties of single-wall BN nanotube, J. Molec.
Struct., 542, 8993, 2001.
96. Lao, J.Y., Wen, J.G., Wang, D.Z., and Ren, Z.F., Synthesis of Amorpohous SiOx
Nanostructures, Department of Physics, Boston College. http://www.physics.bc.edu/
faculty/Ren.html
97. Fagan, S.B., Baierle, R.J., Mota, R., da Silva, A.J.R., and Fazzio, A., Ab initio
calculations for a hypothetical materials: silicon nanotubes, Phys. Rev. B, 61,
99949996, 2000.
98. Spahr, M.E., Stoschitzki-Bitterli, P., Nesper, R., Haas, O., and Novak, P., Vanadium
oxide nanotubes, a new nanostructured redox-active material for the electrochemical
insertion of lithium, J. Electrochem. Soc., 146, 8, 27802783, 1999.
99. Gu, G., Schmid, M., Chiu, P.-W., Minett, A., Fraysse, J., Kim, G.-T., Roth, S., Kozlov,
M., Muoz, E., and Baughman, R.H., V
2
O
5
nanofibre sheet actuators, Nature, 2,
316319, 2003.
100. Hughes, W.L. and Wang, Z.L., Nanobelts as nanocantilevers, Appl. Phys. Lett., 82,
17, 2004.
101. Bong, D.T., Clark, T.D., Granja, J.R., and Ghadiri, M.R., Self-assembling organic
nanotubes, Angew. Chem., 40, 9881011, 2001.
102. Pokropivny, V.V., Non-carbon nanotube (review) 1, synthesis methods, Powder Met-
allurgy Met. Ceram., 40, 485496, 2001.
103. Pokropivny, V.V., Non-carbon nanotube (review) 2, types and structure, Powder
Metallurgy Met. Ceram., 40, 582594, 2001.
104. Pokropivny, V.V., Non-carbon nanotube (review) 3, properties and applications, Pow-
der Metallurgy Met. Ceram., 40, 123135, 2001.
105. Zambov, L., Zambova, A., Cabassi, M., and Mayer, T.S., Template-directed CVD of
dielectric nanotube, Chem. Vapor Deposition, 9, 1, 2633, 2003.
106. Pu, L., Bao, Z., Zou, J., and Feng, D., Individual alumina nanotubes, Angew. Chem.,
113, 8, 15381541, 2001.
107. Sun, X. and Li, Y., Synthesis and characterization of ion-exchangeable titanate
nanotubes, Chem. Eur. J., 9, 22292238, 2003.
108. Tremel, W., Inorganic nanotubes, Angew. Chem. Int. Ed., 38, 15, 21752179, 1999.
109. Chen, Q., Xhou, W., Du, G., and Peng, L.-M., Tritianate nanotubes made via a single
alkali treatment, Advan. Mater., 14, 17, 12081211, 2002.
110. Chang, K.-S., Hernandez, B.A., Fisher, E.R., and Dorhout, P.K., Sol-gel template
synthesis and characterization of PT, PZ and PZT nanotube, J. Korean Chem. Soc.,
46, 3, 242251, 2002.
111. Sha, j., Niu, J., Ma, X., Xu, J., Zhang, X., Yang, Q., and Yang, D., Silicon nanotubes,
Advan. Mater., 14, 17, 12191221, 2002.
112. Mele, E.J. and Kral, P., Electric polarization of heteropolar nanotubes as a geometric
phase, Phys. Rev. Lett., 88, 5, 056803, 2002.
113. Stephan, P., Ajayan, M., Colliex, C., Redlich, Ph., Lambert, J.M., Bernier, P., and
Lefin, P., Doping graphitic and carbon nanotube structures with boron and nitrogen,
Science, 266, 16831685, 1994.
114. Redlich, Ph., Loeffler, J., Ajayan, P.M., Gill, J., Aldinger, F., and Ruhle, M., B-C-N
nanotubes and boron doping of carbon nanotubes, Chem. Phys. Lett., 260, 465470,
1996.
115. Zaporotskova, I.V. and Lebedev, N.G., Some Features of Hydrogenization of Single-
Walled Carbon Nanotubes, Volgograd State University and Institute of Biochemical
Physics, Moscow, Russia.
116. Zaporotskova, I.V. and Lebedev, N.G., The Elastic Modules of Carbon and Boron
Nitride Nanotubes in a Molecular Cluster Model, Volgograd State University and
Institute of Biochemical Physics, Moscow, Russia.
117. Nakhmanson, S.M., Designing Novel Polar Materials through Computer Simulations,
paper presented at Mardi Gras Physics Conference, Louisiana State University, Baton
Rouge LA, February, 2003.
118. Nakhmanson, S.M., Spontaneous Polarization and Piezoelectric Properties of Boron-
Nitride Nanotubes, in Nanotube 2002, Boston College, Boston MA, 2002.
119. Sai, N. and Mele, E.J., Nanotube piezoelectricity, Cond. Mat., 1, 2003.
120. Wang, Z.L., Gao, R.P., Poncharal, P., de Heer, W.A., Dai, Z.R., and Pan, Z.W.,
Mechanical and electrostatic properties of carbon nanotubes and nanowires, Mater.
Sci. Engin. C, 16, 310, 2001.
121. Spinks, G.M., Wallace, G.G., Fifield, L.S., Dalton, L.R., Mazzoldi, A., De Rossi, D.,
Kharyrullin, I.I., and Baughman, R.H., Pneumatic carbon nanotube actuators, Advan.
Mater., 14, 23, 17281732, 2002.
122. Bachtold, A., Hadley, P., Nakanishi, T., and Dekker, C., Logic circuits with carbon
nanotube transistor, Science, 294, 13171320, 2001.
123. White, C.T. and Todorov, T.N., Carbon nanotubes as long ballistic conductors, Nature,
393, 21, 240242, 2001.
124. Liang, W., Bockrath, M., Bozovic, D., Hafner, J.H., Tinkham, M., and Park, H.,
Fabry-Perot interference in a nanotube electron waveguide, Nature, 411, 7, 665669,
2001.
125. Andoa, T., Matsumuraa, H., and Nakanishib, T., Theory of ballistic transport in carbon
nanotubes, Physica B, 323, 4450, 2002.
126. Ebbesen, T.W., Lezec, H.J., Hiura, H., Bennett, J.W., Ghaemi, H.F., and Thio, T.,
Electrical conductivity of individual carbon nanotubes, Nature, 382, 4, 54 56, 1996.
127. C. Schonenberger, A., Bachtold, C., Strunk, et al., Interference and interaction in
multi-wall carbon nanotubes, Appl. Phys. A, 69, 283295, 1999.
128. Frank, S., Poncharal, P., Wang, Z.L., and de Heer, W.A., Nanocapillarity and chem-
istry in carbon nanotubes, Science, 280, 1774, 1998.
129. Roche, S., Triozon, F., Rubio, A., and Mayou, D., Electronic conduction in multi-
walled carbon nanotubes: role of intershell coupling and incommensurability, Phys.
Lett. A, 285, 94100, 2001.
130. Berger, Y., Yi, Z., Wang, L., and de Heer, W.A., Multiwalled carbon nanotubes are
ballistic conductors at room temperature, Appl. Phys. B, 74, 363365, 2002.
131. Radosavljevic, M., Lefebvre, J., and Johnson, A.T., High-field electrical transport and
breakdown in bundles of single-wall carbon nanotubes, Phys. Rev. B, 64, 241307, 2003.
132. Appenxeller, J., Martel, R., and Aavouris, Ph., Phase-coherent transport in ropes of
single-wall carbon nanotubes, Phys. Rev. B, 64, 121404, 2001.
133. Zhao, Y.-P., Wei, B.Q., Ajayan, P.M., Ramanath, G., Lu, T.-M., and Wang, G.-C.,
Frequency-dependent electrical transport in carbon nanotubes, Phys. Rev. B, 64,
201402, 2002.
134. Burke, P.J., An RF circuit model for carbon nanotubes, IEEE Trans. Nanotechnol.,
2, 1, 5558, 2003.
135. Kaiser, A.B., Challis, K.J., McIntosh, G.C., Kim, G.T., Yu, H.Y., Park, J.G., Jhang,
S.H., and Park, Y.W., Frequency and field dependent conductivity of carbon nanotube
networks, Curr. Appl. Phys., 2, 163166, 2002.
136. Qian, D., Wagner, G.J., Liu, W.K., Yu, M.F., and Ruoff, R.S., Mechanics of carbon
nanotubes, ASME, Appl. Mech. Rev., 55, 6, 2002.
137. Qian, D., Liu, W.K., and Ruoff, R.S., Effect of interlayer interaction on the buckling
of multi-walled carbon nanotube, J. Nanosci. Nanotechnol., under review, 2002.
138. Liu, Y.J. and Chen, X.L., Evaluations of the effective materials properties of carbon
nanotube-based composites using a nanoscale representative volume element,
Mechan. Mater., 35, 6981, 2002.
139. Qian, D., Wagner, G.J., and Liu, W.K., Mechanics of carbon nanotubes, ASME, 55,
6, 495533, 2002.
140. Yu, M.-F., Files, B.S., Arepalli, S., and Fuoff, R.S., Tensile loading of ropes of single
wall carbon nanotubes and their mechanical properties, Phys. Rev. Lett., 84, 24, 5552
5555, 2002.
141. Kahn, D. and Lu, J.P., Vibration modes of carbon nanotubes and nanoropes, Phys.
Rev. B, 60, 9, 65356540, 1999.
142. Maarouf, A.A., Kane, C.L., and Mele, E.J., Electronic structure of carbon nanotube
ropes, Phys. Rev. B, 61, 16, 61, 2000.
143. Paulson, S., Helser, A., Nardelli, M.B., Taylor, R.M., II, Falvo, M., Superfine, R.,
and Washburn, S., Tunable resistance of a carbon nanotube-graphite interface, Sci-
ence, 290, 17421744, 2000.
144. Oomman, K., Varghese, D.G., Paulose, M., Ong, K.G., Dickey, E.C., and Grimes,
C.A., Extreme changes in the electrical resistance of titania nanotubes with hydrogen
exposure, Advan. Mater., 15, 78, 624627, 2003.
145. Fischer, J.E., Dai, H., Thess, A., Lee, R., Hanjani, N.M., Dehaas, D.L., and Smalley,
R.E., Metallic resistivity in crystalline ropes of single-wall carbon nanotubes, Phys.
Rev. B, 55, R4921R4924, 1997.
146. Nath, M., Teredesai, P.V., Muthu, D.V.S., Sood, A.K., and Rao, C.N.R., Single-walled
carbon nanotube bundles intercalated with semiconductor nanoparticles, Curr. Sci.,
85, 7, 2003.
147. Tombler, T.W., Zhou, C., Alexseyev, L., Kong, J., Dal, H., Liu, L., Jayanthl, C.S.,
Tang, M., and Wu, S.-Y., Reversible electromechanical characteristics of carbon
nanotubes under local-probe manipulation, Nature, 405, 769772, 2000.
148. Dielectrophoresis tutorial, http://www.ibmm-microtech.co.uk/pages/science/basic.
htm.
149. Vykoukal, J., Sharma, S., Vykoukal, D.M., and Gascoyne, P.R.C., Engineered dielec-
tric microspheres for use in microsystems, University of Texas, M.D. Anderson
Cancer Center, Department of Molecular Pathology.
150. Gascoyne, P.R.C. and Vykoukal, J., Particle separation by dielectrophoresis, Electro-
phoresis, 19731983, 2002.
151. Fuller, C.K., Hamilton, J., Ackler, H., Krulevitch, P., Boser, B., Eldredge, A., Becker,
F., Yang, J., and Gascoyne, P., Microfabricated Multi-Frequency Particle Impedance
Characterization System, fuller14@llne.gov
152. Huang, Y., Yang, J., Wang, X., Becker, F.F., and Gascoyne, P.R.C., The removal of
human breast cancer cells from hematopoietic CDD34+ stem cells by dielectro-
phoretic field-flow fractionalization, J. Hematother. Stem Cell Res., 8, 481490, 1999.
153. Cummings, E.B. and Singh, A.K., Dielectrophoretic Trapping without Embedded
Electrodes, Sandia National Laboratories, Livermore, CA. ebcummi@sandia.gov
154. Cummings, E.B., A comparison of theoretical and experimental electrokinetic and
dielectrophoretic flow fields, AIAA, 20023193.
155. Three Dimensional Nano Manipulation by Using Dielectrophoresis. http://www.mein.
nagoya-u-ac.jp/activity/1999-e/NANO_99E.html
156. Dielectrophoresis, Cell Analysis Ltd. http://www.cell-analysis.com/SciTech/Dielectro/
DielectroBody.html
157. Kral, P. and Shapiro, M., Nanotube electron drag in flowing liquids, Phys. Rev. Lett.,
86, 1, 2001.
158. Ghosh, S., Sood, A.K., and Kumar, N., Carbon nanotube flow sensors, Science, 299, 2003.
159. Shi, D., Lian, J., He, P., Wang, L.M., Xiao, F., Yang, L., Schulz, M.J., and Mastt,
D.B., Plasma coating of carbon nanotubes for enhanced dispersion and interfacial
bonding in polymer composites, Appl. Phys. Lett., 83, 5301, 2003.
160. Bahr, J.L. and Tour, J.M., Covalent chemistry of single-wall carbon nanotubes,
J. Mater. Chem., 12, 19521958, 2002.
161. Zhao, J., Buldum, A., Han, J., and Lu, J.P., Gas molecule adsorption in carbon
nanotubes and nanoparticle bundles, Nanotechnology, 13, 195200, 2002.
162. Bahr, J.L. and Tour, J.M., Covalent chemistry of single-wall carbon nanotubes,
J. Mater. Chem., 12, 19521958, 2002.
163. Appetecchi, G.B., Croce, F., and Scrosati, B., Electrochim. Acta, 40, 991, 1995.
164. Sebhon, S.S., Arora, N., Singh, B., Chandra, A., Protonic polymer gel electrolytes
based on carboxylic acids: ortho and inductive effects, J. Mater. Sci. Lett., 37 10, 2002.
165. NASA Glenn Research. http://www.lerc.nasa.gov/WWW/ictd/content/hilight20020607.
html
166. Ahao, Q., Frogley, M.D., and Wagner, H.D., Direction-sensitive strain-mapping with
carbon nanotube sensors, Composites Sci. Technol., 62, 147150, 2002.
167. Chiarello, G., Maccallini, E., Agostino, R.G., Formoso, V., Cupolillo, A., Pacile, D.,
Colavita, E., Papagno, L., Petaccia, L., Larciprete, R., Lizzit, S., and Goldoni, A.,
Electronic and vibrational excitations in carbon nanotubes, Carbon, 41, 985992, 2003.
168. Shahinpoor, M. and Kim, K.J., Ionic polymer-metal composites: I. Fundamentals,
Smart Mater. Struct., 10, 819833, 2001.
169. Liu, B., Sundqvist, B., Andersson, O., Wagberg, T., Nyeanchi, E.B., Zhu, X.-M., and
Zou, G., Electric resistance of single-walled carbon nanotubes under hydrostatic
pressure, Solid State Commn., 118, 3136, 2001.
170. Peng, S., OKeeffe, J., Wei, C., and Cho, K., Carbon Nanotube Chemical and Mechan-
ical Sensors, paper presented at 3rd International Workshop on Structural Health
Monitoring, September 1214, 2001.
171. Cao, J., Wang, Q., and Dai, H., Electromechanical properties of metallic, quasimetallic
and semiconducting carbon nanotubes under stretching, Phys. Rev. Lett., 90, 15,
157601, 2003.
172. Rochefort, A., Salahub, D., and Avouris, P., The effect of structural distortions on
the electronic structure of carbon nanotubes, Chem. Phys. Lett., 297, 4550, 1998,
2000.
173. Poncharal, P., Wang, Z.L., Ugarte, D., and de Heer, W.A., Electrostatic deflections
and electromechanical resonances of carbon nanotube, Science, 283, 1513, 1999.
174. Varghese, O.K. Kichambre, P.D., Gong, D., Ong, K.G., Dickey, C.C., and Grimes, C.A.,
Gas sensing characteristics of multi-wall carbon nanotube, Sensors Actuators B, 81,
3241, 2001.
175. Ong, K.G. and Grimes, C.A., A resonant printed-circuit sensor for remote query
monitoring of environmental parameters, Smart Mater. Struct., 9, 421428, 2000.
176. Ong, K.G., Zeng, K., and Grimes, C.A., Wireless, passive carbon nanotube-based gas
sensor, IEEE Sensors J., 2, 2, 8288, 2002.
177. Dharap, P., Li, Z., Nagarajaiah, S., and Barrera, E.V., Nanotube film based on single-
wall carbon nanotubes for strain sensing, Nanotechnology, 15, 2004.
178. Schnur, J.N., Lipid tubules: A paradigm for molecularly engineered structures, Sci-
ence, 262, 1669, 1993.
179. Lord, E.A., Helical structures: The geometry of protein helices and nanotubes, Struct.
Chem., 13, 305314, 2002.
180. Terech, P., de Geyer, A., Struth, B., and Talmon, Y., Self-assembled monodisperse
steroid nanotubes in water, Advan. Mater., 14, 7, 495498, 2002.
181. Chung, Y., Kong, M., Modeling of polynucleotide translocation through protein pores
and nanotubes, Electrophoresis, 23, 26972703, 2002.
182. Frusawa, H., Fukagawa, A., Ikeda, Y., Araki, J., Ito, K., John, G., and Shimizu, T.,
Aligning a single-lipid nanotube with moderate stiffness, Angew. Chem., 115, 7678,
2003.
183. Kim, P. and Liever, C.M., Nanotube nanotweezers, Science, 286, 21482150, 1999.
184. Nakayama, Y., Scanning probe microscopy installed with nanotube probes and nan-
otube tweezers, Ultramicroscopy, 91, 4956, 2002.
185. Weissmuller, J., Viswanath, R.N., Kramer, D., Zimmer, P., Wurschum, R., and Gleither,
H., Charge-induced reversible strain in a metal, Science, 300, 312315, 2003.
186. Baughman, R.H. et al., Muscles made from metal, Science, 300, 268269, 2003.
187. Hess, H. and Vogel, V., Molecular shuttles based on motor proteins: active transport
in synthetic environments, Rev. Molecular Biotechnol., 82, 6785, 2001.
188. Schliwa, M. and Woehlke, G., Molecular motor, Nature, 422, 759765, 2003.
189. Ait-Haddou, R. and Herzog, W., Force and motion generation of myosin motors:
muscle contraction, J. Electromyography and Kinesiology, 12, 435445, 2002.
190. VPL actuators. http://npg.nature.com/npg/servlet/Content?data=xml/01_reprints.xml
&style=xml/01_reprints
191. Mavroidis, C., Yarmush, M., Dubey, A., Thronton, A., Nikitszuk, K., Tomassone, S.,
Papadimitrakopoulos, F., and Yurke, B., Protein based nano-machines for space appli-
cations. NiAC Grant Final Report, http://bionano.rutgers.edu/Mavroidis_Final_
Report.pdf
192. Mustarelli, P., Quartarone, E., Tomasi, C., and Magistris, A., New materials for
polymer electrolytes, Solid State Inonics, 135, 8186, 2000.
193. Stephan, A.M., Saito, Y., Muniyandi, N., Renganathan, N.G., Kalyanasundaram, S.,
and Nimma, E.R., Preparation and characterization of PVC/PMMA blend polymer
electrolytes complexed with LiN(CF
3
SO
2
)
2
, Solid State Inonics, 148, 467 473, 2002.
194. Rajendran, S. and Uma, T., Lithium ion conduction in PVCLiBF electrolytes gelled
with PMMA, J. Power Sources, 88, 282285, 2000.
195. Sukeshini, A.M., Nishimoto, A., and Watanabe, M., Transport and electrochemical
characterization of plasticized poly(vinyl chloride) solid electrolytes, Solid State
Inonics, 8688, 385393, 1996.
196. Gupta, P.N. and Singh, K.P., Characterization of H
3
PO
4
based PVA complex system,
Solid State Inonics, 8688, 319323, 1996.
197. Lewis, T.W., Kim, B.C., Spinks, G.M., and Wallace, G.G., Evaluation of solid polymer
electrolytes for use in conducting polymer/nanotube actuators, Proc. SPIE, 3987,
351357, 2000.
198. Lehtinen, T., Sundholm, G., Holmberg, S., Sundholm, F., Bjornbomd, P., and Burselld,
M., Electrochemical characterization of PVDF-based proton conducting membranes
for fuel cells, Electrochimica Acta, 43, 1213, 18811890, 1998.
199. Kaiser, A.B., Flanagan, G.U., Stewart, M., and Beaglehole, D., Heterogeneous model
for conduction in conduction polymers and carbon nanotubes, Synth. Met., 117,
6773, 2001.
200. Smits, J., Wincheski, B., Ingram, J., Watkins, N., and Jordan, J., Controlled deposition
and applied field alignment of single walled carbon nanotubes for CNT device, NASA
Langley, 2003.
201. NASA Ames. http//ettc.usc.edu/ames/nano/visionchip.html
202. Kirikera, G.R., Datta, S., Schulz, M.J., Ghoshal, A., and Sundaresan, M.J., Mimicking
the biological neural system using active fiber continuous sensors and electronic logic
circuits, paper presented at SPIE 11th International Symposium on Smart Structures,
San Diego, March, 2004.
203. Mineta, T., Mitsui, T., Watanabe, Y., Kobayashi, S., Haga, Y., and Esashi, M., Batch
fabricated flat meandering shape memory alloy actuator for active catheter, Sensors
Actuators A, 88, 112120, 2001.
204. Lim, G., Park, K., Sugihara, M., Minami, K., and Esashi, M., Future of active
catheters, Sensors Actuators A, 56, 113121, 1999.
205. Dae-Kyu, K., Yeo-Heung, Y., Kee-Ho, Y., Seong-Cheol, L., Simulation of an Active
Catheter Actuator Using Shape Memory Alloy, Korea Society of Precision Engineer-
ing 2000 Fall Conference, 7275, 2000.
206. Krishna, G.M. and Rajanna, K., Tactile sensor based on piezoelectric resonance, IEEE
Sensors J. A, No, 5, 2004.
207. Sherrit, S., Bao, X., Sigel, D.A., Gradziel, M.J., Askins, S.A., Dolgin, B.P., Bar
Cohen, Y., Characterization of transducers and resonators urder high drive levels,
IEEE International Ultra Sciences Symposium, Atlanta, 2001.
208. Hata, K. Futabal, D.N., Mizuno, K., Namai, T., Yumura, M., and Iijima, S., Water-
assisted highly efficient synthesis of impurity-free single-walled carbon nanotubes,
Science, 306, Novmeber 19, 2004.
209. Cumings, J. and Zettl, A., Localization and nonlinear resistance in telescopically
extended nanotubes, Phys. Rev. Lett. 93, 086801, 14, 2004.
210. Cumings, J., Collins, P.G., and Zettl, A., Peeling and sharpening multiwall nanotubes,
Nature, 406, 586, 2000.
211. Cumings, J. and Zettl, A., Resistance of Telescoping Nanotubes, in Structural and
Electronic Properties of Molecular Nanostructures, Kuzmanyh, H. et a1., Eds., Amer-
ican Institure of Physics, Washington, D.C., 2002.
212. Cumings, J. and Zettl, A., Low-friction nanoscale linear bearing realized from mul-
tiwall carbon nanotubes, Science, 28, Vol 289, July 2000.
213. Lozovik1, Yu.E., Minogin, A.V., and Popov, A.M., Nanomachines Based on Carbon
Nanotubes, Institute of Spectroscopy, Russian Academy of Science, Moscow, Russia.
214. Forro, L., Nanotechnology: Beyond Gedanken Experiments, Department of Physics,
Ecole Polytechnique Federale de Lausanne, Lausanne, Switzerland.
215. Kim, D.H., Chang, K.I., Electron transport in telescoping carbon nanotubes, Physical
Review B 66, 2002.
216. Lozovik. Y.E., Nikolaev, A.G., Popv, A.V., Atomic Scale Design of Carbon Nano-
tubes: the Way to Produce Bolt-and-Nut Pairs, Institute of Spectroscopy of Russia
Academy of Science, Moscow, Russia.
217. Legoas, S.B., Coluci, V.R., Braga, S.F., Coura, P.Z., Dantas, S.O., Galvao, D.S.,
Gigahertz Nanomechanical Oscillators Based on Carbon Nanotubes, TNT, Sept.
1519, 2003, Salamanca Spain.
16
Thermal Properties
and Microstructures
of Polymer
Nanostructured Materials
Joseph H. Koo and Louis A. Pilato
CONTENTS
16.1 Introduction
16.2 Selection of Nanoparticles
16.2.1 Montmorillonite Nanoclays
16.2.2 Carbon Nanofibers (CNFs)
16.2.3 Polyhedral Oligomeric Silsesquioxane (POSS)
16.3 Discussion of Results
16.3.1 TEM Analyses of Nanoparticles
16.3.2 Fire-Retardant Nanocomposite Coatings
16.3.3 Nanostructured Materials for Propulsion Systems
16.3.4 Nanocomposite Rocket Ablative Materials
16.3.5 Nanomodified Carbon/Carbon Composites
16.4 Summary and Conclusions
Problems
Acknowledgments
References
16.1 INTRODUCTION
The utility of introducing inorganic nanomaterials as additives into polymer systems
has resulted in polymer nanostructured materials exhibiting multifunctional, high-
performance polymer characteristics beyond what traditional polymer composites
possess. Multifunctional features attributable to polymer nanocomposites consist of
improved thermal resistance, flame resistance, moisture resistance, decreased per-
meability, charge dissipation, and chemical resistance. Through control and/or alter-
ation of the additive at the nanoscale level, one can maximize property enhancement
of selected polymer systems to meet or exceed the requirements of current military,
aerospace, and commercial applications. The technical approach involves the intro-
duction of nanoparticles into selected polymer matrix systems whereby nanoparticles
may be surface-treated to provide hydrophobic characteristics and enhanced inclu-
sion into the hydrophobic polymer matrix.
The objectives of this chapter are to summarize our research activities in the areas
of (1) developing processes to disperse nanoparticles uniformly in the different types
of polymers, (2) using wide-angle x-ray diffraction (WAXD), transmission electron
microscopy (TEM), and scanning electron microscopy (SEM) techniques to charac-
terize polymer nanocomposite structures, (3) studying the structureproperty rela-
tionships of these types of new materials, and (4) evaluating the thermal performances
of these materials for different applications using established laboratory devices.
Four different processing methods were used to disperse nanoparticles in a
variety of polymers in this study: (1) high shear mixing for liquid resins, (2) three-
roll milling for liquid resins, (3) Brabender-type mixing for high viscosity resins,
and (4) twin screw extrusion for solid polymers. The degree of nanodispersion was
characterized by WAXD, TEM, and SEM analyses. These imaging techniques
allowed us to screen formulations and distinguish compositions that exhibited both
favorable and unfavorable nanodispersed nanoparticle/polymer blends. Furthermore
these analytical techniques facilitated and provided guidelines in the scale-up of
favorable compositions.
Different types of polymers including (1) thermosets, (2) thermoplastics, and
(3) thermoplastic elastomers are presented as examples in this chapter. Three types
of nanoparticles including (1) montmorillonite (MMT) organoclays, (2) carbon
nanofibers (CNFs), and (3) polyhedral oligomeric silsesquioxanes (POSS
) were
used to disperse into the preceding polymers. Selected laboratory devices such as a
cone calorimeter, simulated solid rocket motor (SSRM), and subscale solid rocket
motor were used to evaluate material performance for fire and rocket propulsion
applications.
16.2 SELECTION OF NANOPARTICLES
16.2.1 MONTMORILLONITE NANOCLAYS
Exfoliation of montmorillonite (MMT) clays into polymers has been shown to
increase mechanical properties, barrier performance, and application processing.
1
Achieving exfoliation of organomontmorillonite in various polymer continuous
phases is a function of the surface treatment of the MMT clays and the mixing
efficiency of the dispersing apparatus. Surface treatment of MMT is classically
accomplished with the exchange of inorganic counterions, e.g., sodium, with quater-
nary ammonium ions. The choice of the quaternary ammonium ion is a function of
the hydrophilic and hydrophobic nature of the continuous phase; the hydrophilic and
hydrophobic nature of the interface of MMT is balanced with the hydrophobic and
hydrophilic nature of the continuous phase. Cloisite
30B, a MMT clay from

Southern Clay Products, had been used in several of our application studies and
will be described in detail as an example for nanoclay. It is a surface modified
montmorillonite clay [tallow bishydroxyethyl methyl, T(EOH)
2
M] manufactured by
Southern Clay Products (SCP).
2
The chemical surface structure of Cloisite 30B is
shown in Fig. 16.1. Cloisite 30B is used as an additive for plastics to improve various
physical properties, such as reinforcement, heat distortion temperature, and barrier
properties. It has a ternary ammonium salt modifier at 90 meq/100 g clay and is
off-white in color. Its typical dry-particle size and physical properties are shown
in Table 16.1 and Table 16.2, respectively.
Cloisite 30B in loadings of 5, 10, and 15 wt% was dispersed in a liquid phenolic
resole resin (SC-1008, Borden Chemicals).
36
SC-1008 is the matrix used in the
manufacture of MX 4926, an ablative material for solid rocket nozzles. Mixing
efficiency in the phenolic resin can be achieved with standard paint mixing or mixing
equipment (IKA Works, Inc., Wilmington, NC) that is designed to provide high-
shear mixing if the phenolic resin is in solution. The TEM analyses allowed us to
determine the degree of dispersion/exfoliation of the nanoclay before committing to
a 20-lb production run of these nanoparticle/resin mixtures to make prepregs at
Cytec Engineered Materials (CEM). This proved to be a very cost-effective and
efficient technique for screening different formulations. This nanoclay has also been
used with cyanate ester,
79
thermoplastic elastomer,
10,11
epoxy,
12
and nylon 11
13
in
our research.
16.2.2 CARBON NANOFIBERS (CNFS)
CNFs are forms of vapor-grown carbon fiber, which is a discontinuous graphitic
filament produced in the gas phase from the pyrolysis of hydrocarbons.
1417
In regard
to physical size, performance improvement, and product cost, CNFs complete a
FIGURE 16.1 Chemical structure of Cloisite 30B. T = tallow (~65% C18; ~30% C16; ~5%
C14). Anion: Chloride. MT2EtOH = methyl, tallow, bis-2-hydroxyethyl, quaternary ammonium.
TABLE 16.1
Typical Dry-Particle Sizes for Cloisite 30B (Microns, by Volume)
10% Less Than 50% Less Than 90% Less Than
2 P 6 P 13 P
N
+
CH
2
CH
2
OH
CH
2
CH
2
OH
T CH
3
Where T is tallow (~65% C18; ~30% C16; ~5% C14)
Anion: Chloride
(1) MT2EtOH: methyl, tallow, bis-2-hydroxyethyl, quaternary ammonium
continuum bounded by carbon black, fullerenes, and single- and multi-wall carbon
nanotubes on one end and continuous carbon fibers on the other.
16
CNFs are able
to combine many of the advantages of these other forms of carbon for reinforcement
in engineered polymers. They have transport and mechanical properties that approach
the theoretical values of single-crystal graphite, similar to the fullerenes, but can be
made in high volumes at low cost ultimately lower than conventional carbon
fibers. In equivalent production volumes, a CNF is projected to have a cost compa-
rable to E-glass on a per-pound basis, yet possess properties that far exceed those
of glass and are equal to or exceed those of much more costly commercial carbon
fiber. Maruyama and Alam published an excellent review of carbon nanotubes and
nanofibers in composite materials.
17
CNFs are manufactured by Applied Sciences, Inc./Pyrograf
Products by pyro-
lytic decomposition of methane in the presence of iron-based catalyst particles at
temperatures above 900qC. Pyrograf-III is a patented, very fine, highly graphitic
CNF. Pyrograf-III is available in diameters ranging from 50 to 200 nm and a length
of 50 to 100 Pm. Therefore CNFs are much smaller than conventional continuous
or milled carbon fibers (5 to 10 Pm) but significantly larger than carbon nanotubes
(1 to 10 nm). Compared to PAN and pitch-based carbon fiber, the morphology of
CNFs is unique in that there are far fewer impurities in the filament, providing for
graphitic and turbostatic graphite structures, and the graphene planes are more
preferentially oriented around the axis of the fiber. Consequences of the circum-
ferential orientation of high-purity graphene planes are a lack of cross-linking
between the graphene layers, and a relative lack of active sites on the fiber surface,
making it more resistant to oxidation, and less reactive for bonding to matrix
materials. Also in contrast to carbon fiber derived from PAN or pitch precursors,
CNFs are produced only in a discontinuous form, where the length of the fiber can
be varied from about 100 Pm to several cm, and the diameter is of the order of
100 nm. The two commonly used CNFs in our application studies are PR-19-PS
and PR-24-PS.
312
CNFs exhibit exceptional mechanical and transport properties, allowing them
excellent potential as components for engineering materials. Table 16.3 lists the
properties of vapor-grown carbon fibers, both as-grown and after a graphitizing
heat treatment to 3000qC. Note that due to the difficulty of direct measurements
on the nanofibers, the values shown are measured on vapor-grown fibers that
have been thickened to several microns in diameter.
16
Such fibers consist almost
TABLE 16.2
Physical Properties of Cloisite 30B (Microns, by Volume)
Property Cloisite 30B
Loose bulk density (lb/ft
3
) 14.25
Packed bulk density (lb/ft
3
) 22.71
Specific gravity 1.98
d
001
() 18.5
exclusively of chemical vapor deposition (CVD) carbon, which is less graphitic and
more defective than the catalytically grown carbon core that constitutes the CNF.
Thus the properties listed in the table represent an estimate of the properties of CNFs.
One of the goals for CNF broad utility is to provide mechanical reinforcement
comparable to that achieved with continuous carbon ber at a price that approaches
that of glass ber reinforcement and with low-cost composite fabrication methods
such as injection molding. Theoretical models
16
suggest that reinforcement by
discontinuous bers such as CNFs can closely approach that of continuous bers
as long as the aspect ratio of the bers is high and the alignment is good. Work is
ongoing to improve the mechanical benets of CNFs through ber surface modi-
cation to provide physical or chemical bonding to the matrix. Such modications
have resulted in strength and modulus improvements of four to six times the values
of the neat resin; however, these values are still a modest fraction of what may be
anticipated from idealized bermatrix interphase and alignment of the bers within
the matrix. The more immediate opportunities for use in structural composites lie
in the prospect for modifying the properties of the matrix material. For example,
use of small volume loadings of CNFs in epoxy may allow for improvement of
interlaminar shear strength of PAN or pitch-based composites. The CNF additives
to berglass composites could provide benets to a suite of properties, including
thermal and electrical conductivity, coefcient of thermal expansion, and mechan-
ical properties, as suggested by the data in Table 16.4.
16
Figure 16.2 is an SEM
TABLE 16.3
Properties of CNFs
Property (Units) As Grown Heat Treated
Tensile strength (GPa) 2.7 7.0
Tensile modulus (GPa) 400 600
Ultimate strain (%) 1.5 0.5
Density (g/cc) 1.8 2.1
Electrical resistivity (-cm) 1000 55
Thermal conductivity (W/m-K) 20 1950
TABLE 16.4
Thermoset Polyester/Pyrograf-III Composite Properties
Fiber Content (Wt%)
Tensile Strength
(MPa)
Tensile Modulus
(GPa)
Electrical Resistivity
(ohm-cm)
17% PR-19 51.5 4.55 3.2
17% PR-19, OX 47.4 4.55 7.1
5% PR-19 in 10%
1
/4" glass 44.1 11.52 5.0
5% PR-19, OX in 10%
1
/4" glass 33.8 8.92 7.0
micrograph of a CNF bundle.
16
Figure 16.3 is a TEM micrograph of an individual
CNF with a hollow core in two distinct regions: catalytic and deposited.
16
Recently Glasgow et al.
18
demonstrated that the achievement of significant
mechanical reinforcement in CNF composites requires high fiber loadings and is
somewhat dependent on generating an appropriate interphase between the CNF and
the matrix. Novel surface treatments under development have yielded good improve-
ments in the tensile modulus and strength of CNF-reinforced polypropylene. Adding
surface functional groups, particularly oxygen groups, also demonstrated benefits
for interphase development. Carboxyl and phenolic groups contributing to a total
surface oxygen concentration in the range from 5 to 20 atom percent have been
added to CNF used to fabricate epoxy polymer matrix composites, providing
FIGURE 16.2 SEM micrograph of CNF bundles.
FIGURE 16.3 TEM micrograph of individual CNF showing hollow core and two distinct
regions: catalytic and deposited.
improved flexural strength and modulus. The effect of similarly functionalized CNF
in bismaleimide (BMI) polymer matrix composites also shows promise. Data for
polypropylene, epoxy, and BMI/CNF-reinforced composites indicate that higher
fiber volume loadings will find a role in structural composite markets as price and
availability improve.
16.2.3 POLYHEDRAL OLIGOMERIC SILSESQUIOXANE (POSS)
Representing a merger between chemical and filler technologies, POSS nanostruc-
tured materials can be used as multifunctional polymer additives, acting simulta-
neously as molecular level reinforcements, processing aids, and flame retardants.
POSS nanostructured materials have two unique structural features: (1) the chemical
composition is a hybrid, intermediate (RSiO
1.5
) between those of silica (SiO
2
) and
silicones (R
2
SiO); (2) the POSS molecules are nanoscopic in size, ranging from
approximately 1 to 3 nm. POSS molecules have multifunctional properties and an
inorganic framework of silicone oxide. It has been proven that this is ceramic-like
in nature, especially in severe environments. The other key factor for POSS property
enhancement is the length scale effect.
Lee
19
followed the approach in using POSS as an additive alloyed into epoxy
systems and as a copolymer compatibilizer. A POSS-alloyed epoxy sample prepa-
ration procedure was developed. The epoxy was preheated for an hour, and then
POSS in different weight percentages was added. The two epoxies used were the
DER 332 system, which is the difunctional epoxy of the DGEBA variety and the
tetrafunctional epoxy of 4,4c-methylene bisaniline. The trisilanolphenyl-POSS was
selected for his work. A very low amount of loading (0.20 to 1.00 wt%) of the POSS
compound was used. At 0.20 wt% of POSS, he reported a significant increase in T
g
from 84.2 to 92.1 as shown in Table 16.5. For the tetrafunctional epoxy also
mixed with Jeffamine, the same increase in the glass transition temperature was
observed.
19
When a Dytek A curing agent was used on the DER-332 and 4,4
TABLE 16.5
Summary of DMA and DSC Thermal Data for Jeffamine D-230 Cured
DER-332 Epoxy System
POSS (wt%) Tg DMTA (T)
*
Tg DSC (T)
**
0% 84.2 (12) 86.6 (6)
0.20% 92.1 (10) 90.1 (5)
0.40% 91.2 (9) 91.4 (5)
0.60% 90.6 (9) 94.4 (6)
0.80% 91.6 (10) 91.1 (5)
1.00% 91.7 (12) 91.7 (7)
*
Tg is determined by using tan G curve; T is the full width at half the tan G peak.
**
T is endset temperatureonset temperature.
methylenebis (N-N-diglycidylaniline) epoxy system, he did not observe any change
in DSC. The particular diamine used to cure the system is critical.
Hybrid Plastics SO1458 trisilanolphenyl-POSS (C
42
H
38
O
12
Si
7
) has been used
in several of our studies. It is the material of choice for use with phenolic,
36
epoxy,
12
cyanate ester,
79
and thermoplastic elastomer.
10,11
This was also used by Lee.
19
Figure
16.4 shows the chemical structure of the trisilanolphenyl-POSS. POSS is currently
very expensive at the research and development material level because it is a new
nanoparticle with low-production volume.
16.3 DISCUSSION OF RESULTS
The challenges to form polymer nanocomposites are (1) choice of nanoparticles
involving an interfacial interaction or compatibility with the polymer matrix, (2)
proper processing techniques to uniformly disperse and distribute the nanoparticles
or aggregates within the polymer matrix, and (3) selective analytical tools to char-
acterize the morphology of the nanocomposites. We selected four application areas
as examples for this chapter: (1) fire retardant nanocomposite coatings, (2) nano-
structured materials for rocket propulsion systems with improved thermal insulation,
(3) nanocomposite rocket ablative materials, and (4) nanomodified carboncarbon
composites. These examples are used to illustrate the processing techniques to form
the resulting nanocomposite system, the analytical and imaging techniques used to
characterize the nanocomposites, and the thermal property and performance char-
acteristics of these nanocomposites for their specific applications.
16.3.1 TEM ANALYSES OF NANOPARTICLES
Three types of nanoparticles were dissolved in 0.5% isopropanol (IPA) solvent
overnight to create specimens for TEM analyses. Figure 16.5 shows that the Cloisite
30B nanoclay layers are in an intercalated state and the scale bar is 500 nm. Figure
16.6 shows the PR-24-PS CNFs are entangled and the scale bar is 1 Pm. Figure
16.7 shows trisilanolphenyl-POSS-SO1458 (C
48
H
38
O
12
Si
7
) showing micron and
nanometer particles in the system and the scale bar is 1 Pm. Neat nanoparticles
dissolved in isopropanol solvent exhibit particle intercalation, entanglement, and a
wide assortment of micron and nanometer particles.
FIGURE16.4 Chemical structure of trisilanolphenyl-POSS.
O
O
O
O
O
O
O
OH
OH
OH
O
O
Si
Si
Si
Si
Si
Si Si
R
R
R
R
R
R
R
R = phenyl
SO1458
TriSilanolPhenyl-POSS
C
42
H
38
O
12
Si
7
MW: 931.34 g/mole
Soluble: tetrahydrofuran, chloroform, ethanol
Insoluble: hexane
Appearance: white powder
FIGURE 16.5 TEM micrograph of Cloisite 30B nanoclays dissolved in IPA solvent. Scale
bar = 500 nm.
FIGURE 16.6 TEM image of PR-24-PS CNF dissolved in IPA solvent. Scale bar = 1 Pm.
FIGURE 16.7 TEM image of trisilanolphenyl-POSS-SO1458 (C
48
H
38
O
12
Si
7
) dissolved in
IPA solvent. Scale bar = 1 Pm.
16.3.2 FIRE-RETARDANT NANOCOMPOSITE COATINGS
Vinyl acetateacrylic copolymer (PVA 5174), ammonium polyphosphate (APP),
melamine and dipentaerythritol supplied by StanChem (Hartford, CT), and nanoclay
(Cloisite Na
+
) supplied by Southern Clay Products were used as received. The
nanoclay was dispersed in deionized water with Dispermat (high-shear mixer) at
high speed for about 1 h. The slurry was allowed to stay at room temperature
overnight. The slurry was then mixed with PVA 5174 and/or intumescent compo-
nents (APP, dipentaerythritol, and melamine) for about 2 h for subsequent coating.
Douglas fir plywood sized 10 by 10 by 1.8 cm was used as the substrate for all test
specimens. A 5-ml plastic syringe was used to measure and place the amount of
coating on the sample. It was evenly spread (visual monitoring only) with a brush
presaturated with the coating. The coating thickness was monitored with a digital
caliper for a dry thickness of 0.28 to 0.30 mm (0.011 to 0.012 in.). The edges were
then fully coated and the coating was allowed to dry at room temperature for about
24 h before testing. Samples were exposed to a mass loss calorimeter under a heat
flux of 50 kW/m
2
for 15 min. The calorimetry data reported here represent the
average of three replicates.
20
Figure 16.8 shows the heat release rates of PVAclay fire-retardant (FR) coating
systems with different clay loading at a heat flux of 50 kW/m
2
. It is very clear that
nanoclay plays a key role in reducing the flammability of these coating systems.
With only 10% clay, the peak high release rate (PHRR) was dramatically reduced
by almost 50%. With more clay applied, the PHRR continued to drop. Actually with
FIGURE 16.8 Comparison of HRR values for PVAclay fire-retardant coatings at 50 kW/m
2
.
0
20
40
60
80
100
120
140
0 200 400 600 800
PVA-Clay10
PVA-Clay20
PVA-Clay30
PVA-Clay50
PVA-Clay60
H
R
R

(
k
W
/
m
2
)
Time (s)
50% and 60% clay loading, the testing specimen could not be ignited throughout
the 15-min test. Figure 16.9 shows the mass loss of the PVAclay fire-retardant
coating systems. It can be seen that with the addition of clay, the residual mass has
been raised from 32% to about 70%.
A discernable effect of the MMT clay on the HRR data is supported by Fig.
16.10. Without any MMT clay, the wood was severely burnt. As a result, a higher
HRR was observed throughout the test. However, with the MMT clay loading up to
30%, it is clear that a uniform ceramic layer has been formed on the top of the wood
block, which protects the wood from heat throughout the test. Therefore a lower
HRR has been obtained.
FIGURE 16.9 Comparison of residual mass for PVAclay fire-retardant coatings at 50
kW/m
2
.
FIGURE 16.10 Digital photos of combustion residues of wood blocks coated with PVA 5174
(left), PVA3% MMT clay (center), and PVA30% MMT clay (right).
0
20
40
60
80
100
0 200 400 600 800
PVA-Clay10
PVA-Clay20
PVA-Clay30
PVA-Clay50
PVA-Clay60
PVA5174
W
t
.
%
Time (s)
It has been shown that incorporation of MMT clay in the PVA coating systems
significantly improved its fire retardant properties at clay loading of 30% or higher.
A thin section of the post-test specimen of PVAclay nanocomposite fire-retardant
coating system with 30% clay loading was studied by scanning electron microscopy
(SEM) and transmission electron microscopy (TEM). The SEM micrographs in Fig.
16.11 show a post-test specimen of PVA30% clay loading at low and high magni-
fications. It is evident that the MMT clay formed protective layers underneath the
charred top surface of the coating. The TEM micrographs of PVA30% clay post-
test residue are shown in Fig. 16.12. Distinct nanoclay layers in the post-test ash
residue of PVA30% clay loading specimen were observed.
16.3.3 NANOSTRUCTURED MATERIALS FOR PROPULSION SYSTEMS
Polymer chopped fiber-filled systems are used as insulative materials in rocket propul-
sion systems. This research program is aimed to develop new classes of nanostructured
materials that are lighter and have better erosion and insulation characteristics than
current insulative materials (Kevlar
-filled EPDM rubber). The TPSiV X1180 ther-

moplastic elastomer (TPE) is a polyamide-based vulcanized silicone thermoplastic resin
manufactured by Dow Corning and was selected as a potential replacement material.
10,11
Its typical uses include profiles for automotive fuel and vapor line covers, brake hose
covers, and industrial applications involving extruded profiles exposed to harsh envi-
ronments. Table 16.6 shows the chemical compositions of three thermoplastic elas-
tomer/ nanoparticle blends produced by twin screw extrusion.
FIGURE 16.11 SEM micrographs of post-test PVA30% clay specimen showing ash and
unburned PVA matrix (top left), layers of ash/MMT clay (top right), and different views of
ash layers (bottom left and bottom right).
TEM analyses were conducted on the aforementioned three blends. Figure 16.13
shows the TEM images of the Dow Corning polyamide silicone TPSiV X1180 in
two separate phases; the dark color is the silicone phase and the light color is the
polyamide phase. Figure 16.14 shows the TEM images of the 7.5 wt% Cloisite 30B
92.5 wt% TPSiV X1180 blend. We speculate that the Cloisite 30B nanoclays are
dispersed only in the polyamide phase, since the silicone phase is already cross-
linked. TEM images of the 15 wt% PR-24-PS CNF85 wt% TPSiV X1180 blend
(Fig. 16.15) and 10% Ph
12
T
12
POSS 90 wt% TPSiV X1180 blend (Fig. 16.16)
FIGURE 16.12 TEM micrographs of PVAclay 30% post-test ash showing layers of clay
residues at low (top) and high magnification (bottom) on grids supported by carbon film.
TABLE 16.6
Thermoplastic Elastomer/Nanoparticle Blends
Material TPE (wt%) Nanoparticle (wt%)
1 TPSiV X1180 (90%) Ph
12
T
12
-POSS (10%)
2 TPSiV X1180 (92.5%) Cloisite 30B (7.5%)
3 TPSiV X1180 (85%) PR-24-PS CNF (15%)
Enlargement
Overview
PVA-clay30
TEM images of ash/PVA
matrix on grids supported
by carbon film
showed similar results. It was determined that the TPSiV X1180 TPE is difficult to
process with any of the three selected nanoparticles. None was compatible with the
resin system.
The preceding three blends with the baseline Kevlar-filled EPDM rubber were
tested for ablation resistance using a subscale solid rocket motor. Figure 16.17
shows the ablation rates of the materials at low, medium, and high Mach number
regions inside the rocket motor.
10
All three TPEs were out-performed by the
baseline material. These observations clearly demonstrate that if the nanoparticles
are poorly dispersed in the polymer matrix, no thermal performance improvement
can be expected.
FIGURE 16.13 TEM micrographs of polyamide silicone TPSiV X1180 (scale bar = 500 nm).
FIGURE 16.14 TEM micrographs of 92.5% polyamide silicone TPSiV X1180 with 7.5%
Cloisite 30B (scale bar = 500 nm).
16.3.4 NANOCOMPOSITE ROCKET ABLATIVE MATERIALS
This program is aimed at developing nanocomposite rocket ablative materials for solid
rocket nozzles.
36
The candidate materials were evaluated for dispersion uniformity
using WAXD and TEM prior to full ablation testing. Borden SC-1008 resole phenolic
in isopropanol (IPA) is the baseline resin system. Our first attempts were to modify
this resin by the incorporation of MMT organoclays. The surface treatment on these
clays (usually with quaternary ammonium ions) is critical for blend compatibilization
and ease of dispersion of the surface-treated clay into the hydrophobic phenolic resin
matrix. This treatment is a quaternary ammonium ion (quat). These blends were
dispersed using high-shear, nonsparking paint mixing equipment. Neat resin castings
(without fiber reinforcement) were made for WAXD and TEM analyses.
FIGURE 16.15 TEM micrographs of 85% polyamide silicone TPSiV X1180 with 15% PR-
24-PS CNF (scale bar = 1 Pm).
FIGURE 16.16 TEM micrographs of 90% polyamide silicon TPSiV X1180 with 10% Ph
12
T
12
-POSS (scale bars = 1 Pm and 500 nm).
MX-4926 is a rayon carbon fabric impregnated with SC-1008 containing carbon
black (CB) particles. Based on TEM analyses, we concluded that CB particles in the
baseline MX-4926 caused interference of the dispersions of all nanoparticles in the SC-
1008. As a result, we eliminated CB in our subsequent blending experiments. Cloisite
30B in loadings of 5, 10, and 15 wt% was dispersed in the SC-1008. Figure 16.18
FIGURE 16.17 Ablation rates of polyamide silicone TPSiV X1180 with Cloisite 30B,
Ph
12
T
12
-POSS, and PR-24-PS CNF as compared to EPDM/Kevlar in low, medium, and high
Mach number regions inside a rocket motor.
FIGURE 16.18 WAXD of 5, 10, and 15 wt% Cloisite 30B in 95, 90, and 85 wt% SC-1008
phenolic resin, respectively.
0
2
4
6
8
10
12
14
16
18
E
P
D
M
/
K
e
v
l
a
r
T
P
S
i
V
1
1
8
0
/
3
0
B
T
P
S
i
V
1
1
8
0
/
P
h
1
2
T
1
2
T
P
S
i
V
1
1
8
0
/
P
R
-
2
4
-
P
S
Materials
A
b
l
a
t
i
o
n

r
a
t
e

(
m
i
l
s
/
s
)
Low mach
Med. mach
High mach
0
500
1000
1500
2000
5% 30B, 4.1% ash
10% 30B, 8.2% ash
15% 30B, 12.9% ash
2 Theta
41.8 ang
36.6 ang
16.4 ang
I
n
t
e
n
s
i
t
y
,

c
o
u
n
t
s

p
e
r

s
e
c
o
n
d
2 4 6 8 10
shows the WAXDs of the three nanoclay loadings in SC-1008 and indicates good
dispersibility of the nanoclay into SC-1008. Figure 16.19, Fig. 16.20, and Fig. 16.21
are TEM images of the 5 wt% Cloisite 30B in 95 wt% SC-1008 (Fig. 16.19), 10
wt% Cloisite 30B in 90 wt% SC-1008 (Fig. 16.20), and 15 wt% Cloisite 30B in
85 wt% SC-1008 (Fig. 16.21). The TEM images indicate intercalation and not
FIGURE 16.19 TEM images of 5 wt% Cloisite 30B in 95 wt% SC-1008 (scale bar = 1 Pm).
FIGURE 16.20 TEM images of 10 wt% Cloisite 30B in 90 wt% SC-1008 (scale bar = 1 Pm).
FIGURE 16.21 TEM images of 15 wt% Cloisite 30B in 85 wt% SC-1008 (scale bar = 500
nm).
exfoliation of the nanoclay in the resin system. The TEM analyses assisted in
determining the degree of dispersion/exfoliation of the nanoclay before committing
to a 20-lb run of these nanoparticle/resin mixtures to fabricate prepregs at Cytec
Engineered Materials (CEM). This procedure is a very cost-effective and efficient
technique for screening different formulations.
We selected SC-1008 with 5, 10, and 15 wt% Cloisite 30B to replace 15 wt%
of carbon black in the original MX-4926 formulation. CEM prepared three versions
of MX-4926 alternates designated MX-4926 ALT 5%, MX-4926 ALT 10%, and
MX-4926 ALT 15%. Three loadings of PR-24-PS in 20, 24, and 28 wt% were
dispersed in SC-1008 without the rayon carbon fiber reinforcement. Three loadings
of trisilanolphenyl-POSS in 2, 6, and 10 wt% were also dispersed in SC-1008. The
POSSSC-1008 mixture was used with the rayon carbon fabric to produce prepregs.
Table 16.7 shows the chemical compositions for the laminates used for ablation
testing. Figure 16.22 compares the densities of three types of nanocomposite rocket
ablative materials (NRAMs) with nanoclay, CNF, and POSS at various loading levels
of nanoparticles. All CNF NRAMs and POSS NRAMs had densities lower than
MX-4926.
A small-scale supersonic liquid-fueled rocket motor (SSRM) burning kerosene
and oxygen was used to study the ablation and insulation characteristics of the abla-
tives. It was demonstrated as a cost-effective laboratory device to evaluate different
TABLE 16.7
Specimen Configuration for SSRM Laminates Fabrication
Material ID
Density
(g/cc)
Rayon Carbon Fiber
Reinforcement (wt%)
Resin SC-1008
Phenolic (wt%) Filler (wt%)
MX-4926
(Control)
1.44 50 35 15 carbon black (CB)
MX-4926
ALT Clay 5%
1.42 50 47.5 2.5 Cloisite 30B
MX-4926
ALT Clay 10%
1.43 50 45 5 Cloisite 30B
MX-4926 ALT
Clay 15%
1.43 50 42.5 7.5 Cloisite 30B
PR-24-PS
20%/SC-1008
1.35 None 80 20 PR-24-PS CNF
PR-24-PS
24%/SC-1008
1.38 None 76 24 PR-24-PS CNF
PR-24-PS
28%/SC-1008
1.41 None 72 28 PR-24-PS CNF
MX4926 ALT
SO-1458 2%
1.41 50 49 1 Trisilanolphenyl-POSS
MX4926 ALT
SO-1458 6%
MX4926 ALT
SO-1458 10%
ablatives under identical conditions for initial material screening and development.
36
Figure 16.23 shows the peak erosion levels of MX-4926 and MX-4926 ALTs at the
three heat fluxes with Al
2
O
3
particles. As expected, peak erosion decreased as heat
flux decreased. At high heat flux (1000 Btu/ft
2
-sec) the MX-4926 and MX-4926 ALTs
showed more difference in peak erosion performance, while the medium- and low-
heat flux conditions showed little difference in peak erosion. MX-4926 ALT 5 and
FIGURE 16.22 Densities of MX-4926 and NRAMs with different nanoparticle loadings.
FIGURE 16.23 Peak erosions of nanoclay NRAMs at 250, 625, and 1000 Btu/ft
2
-s with
particles.
1.2
1.25
1.3
1.35
1.4
1.45
1.5
1.55
1.6
0 5 10 15 20 25 30
D
e
n
s
i
t
y
,

g
/
c
m
3
Nanoparticle loading, wt. %
PR-24
(HE)
2
MT
SO-1458
MX-4926
0.05
0.1
0.15
0.2
0.25
0.3
0.35
200 400 600 800 1000 1200
Peak erosion depth for
nanoclay NRAM composites
Heat flux, (Btu/ft
2
.
sec)
MX-4926
30B, 5%
30B, 10%
30B, 15%
P
e
a
k

e
r
o
s
i
o
n

(
i
n
c
h
e
s
)
10% Cloisite 30B samples were not as erosion resistant as the control MX-4926. The
MX-4926 ALT 15% showed better erosion characteristics than MX-4926 by 14%.
Figure 16.24 shows all nanoclay NRAM composites exhibited lower maximum
backside heat-soaked temperature rises than MX-4926 by 28%. Surface temperature
of MX-4926 ALTs were lower than MX-4926.
Figure 16.25 shows both PR-19-PS and PR-24-PS CNF NRAMs (without rayon
carbon fabric) have significantly less peak erosion than MX-4926 at 1000 Btu/ft
2
-s
FIGURE 16.24 Maximum backside heat-soaked temperature rises for nanoclay NRAMs at
250, 625, and 1000 Btu/ft
2
-s with particles.
FIGURE 16.25 Peak erosion depths of nanoclay and CNF NRAMs at 250, 625, and 1000
Btu/ ft
2
-s with particles.
60
70
80
90
100
110
200 400 600 800 1000 1200
Backside heat soak temperature change
for NRAM nanoclay composites
Heat flux (Btu/ft
2
.
sec)
MX-4926
30B, 5%
30B, 10%
30B, 15%
T
e
m
p
e
r
a
t
u
r
e

r
i
s
e

(
C
)
0.05
0.1
0.15
0.2
0.25
0.3
0.35
200 400 600 800 1000 1200
Peak erosion depth for
NRAM composites
Heat flux (Btu/ft
2
.
sec)
30B, 5%
MX-4926
PR-24
PR-19
P
e
a
k

e
r
o
s
i
o
n

(
i
n
c
h
e
s
)
by about 42%. Figure 16.26 shows both PR-19-PS and PR-24-PS CNFs also have
substantial lower maximum backside heat-soaked temperature rises than MX-4926
by about 68% at all three levels of heat fluxes. Surface temperatures of the CNF
samples were hotter than the MX-4926 and MX-4926 ALTs. This suggests we may
have produced better radial heat transfer than axial heat transfer supported by the
glowing heat of the surface observed during materials testing.
Figure 16.27 shows the ablation rates of MX-4926 and three groups of NRAMs: clay
NRAM [(HE)
2
MT nanoclay], CNF NRAM (PR-24), and POSS NRAM (SO-1458). The
FIGURE 16.26 Maximum backside heat-soaked temperature rises at 250, 625, and 1000
Btu/ft
2
-s with particles.
FIGURE 16.27 Ablation rates of MX-4926 and NRAMs with different types of nanoparticles
at various loading levels.
50
60
70
80
90
100
110
200 400 600 800 1000 1200
Backside heat soak temperature change
for NRAM composites
Heat flux (Btu/ft
2
.
sec)
MX-4926
30B, 5%
PR-19
PR-24
T
e
m
p
e
r
a
t
u
r
e

r
i
s
e

(
C
)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 5 10 15 20 25 30
A
b
l
a
t
i
o
n

r
a
t
e
,

m
m
/
s
e
c
Nanoparticle loading, wt %
(HE)
2
MT
PR-24
SO-1458
MX-4926
ablation rate of MX-4926 is about 0.4 mm/s. For the clay NRAM group, only the
7.5 wt% clay NRAM has a lower ablation rate than MX-4926. For the CNF NRAM
group, all three loadings have lower ablation rates than MX-4926 with 28% CNF
NRAM being the lowest. For the POSS NRAM group, all three loadings have
lower ablation rates than MX-4926 with 5% POSS NRAM being the lowest. The
loadings of POSS are 1, 3, and 5 wt%, the lowest of the three NRAM groups. The
loadings of clay are 2.5, 5, and 7.5 wt%, the medium of the three NRAM groups.
The loadings of CNF are 20, 24, and 28 wt% without the rayon carbon reinforce-
ments, the highest of the three NRAM groups. Figure 16.28 shows the residual
masses of MX-4926 and all the NRAMs. The residual mass of MX-4926 is about
92%. The POSS NRAM group has the most residual mass, about 93% for all three
loadings. The clay NRAM group has about the same residual mass as the MX-
4926. The CNF NRAM group has about 86 to 88 wt% residual mass, the lowest
of all.
Figure 16.29 shows the maximum backside heat-soaked temperature rises of
MX-4926 and the NRAMs. All NRAMs exhibited lower maximum backside heat-
soaked temperature rises than MX-4926. The backside heat-soaked temperature rise
of MX-4926 was about 106qC. It is obvious that the CNF-NRAM group had sub-
stantial lower maximum backside heat-soaked temperature rises than MX-4926,
from 54q to 72qC. The POSS-NRAM group has the second lowest than MX-4926,
from 75q to 86qC. The clay NRAM group had the third lowest from MX-4926, from
82q to 98qC.
An IR pyrometer was used to measure the surface temperatures of all materials
during SSRM firings. Figure 16.30 shows the surface temperatures of MX-4926 and
the NRAMs. Surface temperatures of MX-4926 were about 1700qC. Surface temper-
atures of the CNF NRAM samples were hotter than the MX-4926, clay NRAMs, and
POSS NRAMs. This suggests better radial heat transfer than axial heat transfer,
supported by the glowing heat of the surface observed during material testing. This
phenomenon was observed by other researchers
21
and needs further study. The
FIGURE 16.28 Residual masses of MX-4926 and NRAMs with different types of nanopar-
ticles at various loading levels.
80
85
90
95
100
0 5 10 15 20 25 30
R
e
s
i
d
u
a
l

m
a
s
s
,

w
t
.

%
PR-24
SO-1458
(HE)
2
MT
MX-4926
surface temperatures of clay NRAMs and POSS NRAMs are lower than MX-4926.
There is essentially no effect of the amount of nanoclay in the clay NRAMs. The
amount of POSS in the POSS NRAMs has a significant effect on the surface
temperature of the POSS NRAMs.
16.3.5 NANOMODIFIED CARBON/CARBON COMPOSITES
The objective of this materials program is to develop an improved carbon/carbon
(C/C) composite with enhanced thermo-oxidative resistance at intermediate temper-
atures (700q to 1200qF).
79
We proposed that a nanophase be introduced into the C/C
FIGURE 16.29 Backside temperature rises of MX-4926 and NRAMs with different types
of nanoparticles at various loading levels.
FIGURE 16.30 Surface temperatures of MX-4926 and NRAMs with different types of nano-
particles at various loading levels.
50
60
70
80
90
100
110
0 5 10 15 20 25 30
B
a
c
k
s
i
d
e

t
e
m
p
e
r
a
t
u
r
e

r
i
s
e

(
C
)
PR-24
(HE)
2
MT
SO-1458
MX-4926
1000
1200
1400
1600
1800
2000
0 5 10 15 20 25 30
S
u
r
f
a
c
e

t
e
m
p
e
r
a
t
u
r
e
,

C
MX-4926
(HE)
2
MT
PR-24
SO-1458
composites (CCCs), prior to cure, to provide improved and maintained mechanical
strength by preventing oxidation of the composites. The candidate materials were
evaluated for dispersion using WAXD and TEM prior to full scale-up. Baseline
phenolic resin 134A and cyanate ester resins PT-15 and PT-30 were used as matrix
resins. Lab scale dispersion of (1) PT-30 cyanate ester (CE) resin with different wt%
POSS, nanoclays, and CNF, (2) PT-15 CE resin with different wt% of POSS and
nanoclays, and (3) 134A phenolic resole resin with different wt% POSS and nanoclays
were conducted. The morphology of selective resin/nanoparticle systems was char-
acterized using TEM and SEM analyses. Detailed processing and characterization of
PT-30/nanoparticle, PT-15/nanoparticle, 134A/nanoparticle systems are reported else-
where.
9
A brief discussion of PT-30/nanoparticle and PT-15/nanoparticle systems is
included in this section.
Cloisite 30B was first dispersed in THF before blending with the PT-30 resin.
Cloisite 30B was uniformly dispersed in PT-30/30B THF system. At higher magni-
fication, Cloisite 30B was in an intercalated state in the PT-30 resin, and larger clay
tactoids were observed than when PT-15 was mixed with Cloisite 30B and cured in
the same manner (Fig. 16.31).
Different weight percents (wt%) of the seven POSS chemicals were blended
with PT-30 for a total of 15 blends using a lab scale high-shear mixer. Appearance
in terms of transparency, translucency, or opaqueness was examined during mixing,
before and after curing, and was recorded for all blends and densities. Based on
visual observation, trisilanolphenyl-POSS is the only potential POSS compound that
works well with PT-30 using a direct melt blending process. All other blends were
either opaque or translucent and large, poorly dispersed particles were observed.
Phase separation had clearly taken place.
Selective candidates were cured for TEM analyses. PT-15 and PT-30 were cured
thermally. The SO1458 POSS particles, when directly blended into PT-30 resin,
showed very few undissolved POSS particles in the resin matrix, as shown in Fig.
16.32. Some molecular dispersion in the PT-30/SO1458 POSS (95/5) was achieved,
as shown in Fig. 16.32. Significant Si was detected in the resin matrix where no
phase separation could be detected. SO1458 is the preferred POSS system for
FIGURE 16.31 Higher magnification of PT-30/Cloisite 30B (97.5/2.5) THF showing nano-
clay in an intercalated state in the PT-30 cyanate ester resin where scale bars are 500 nm
(left), 200 nm (center), and 100 nm (right).
F
I
G
U
R
E

1
6
.
3
2
T
E
M

m
i
c
r
o
g
r
a
p
h
s

o
f

P
T
-
3
0
/
S
O
1
4
5
8

P
O
S
S

(
9
5
/
5
)

a
t

h
i
g
h

m
a
g
n
i
f
i
c
a
t
i
o
n

s
h
o
w
i
n
g

m
o
l
e
c
u
l
a
r

d
i
s
p
e
r
s
i
o
n

o
f

S
O
1
4
5
8

P
O
S
S

w
i
t
h

s
o
m
e

P
O
S
S

p
a
r
t
i
c
l
e
s

i
n

t
h
e

P
T
-
3
0

c
y
a
n
a
t
e

e
s
t
e
r
.

S
i
g
n
i
f
i
c
a
n
t

S
i

w
a
s

d
e
t
e
c
t
e
d

i
n

t
h
e

r
e
s
i
n

m
a
t
r
i
x
.
S
i
O
C
C
u
C
u
C
u
S
i
O
C
C
u
C
u
C
u
P
a
r
t
i
c
l
e

2
F
i
l
e

#
1
9
2
F
i
l
e

#
1
9
4
M
a
t
r
i
x
S
i
O
C
C
u
C
u
C
u
S
i
O
C
C
u
C
u
P
a
r
t
i
c
l
e

1
P
a
r
t
i
c
l
e

3
F
i
l
e

#
1
9
1
F
i
l
e

#
1
9
3
M
g A
l
C
u
0
2
0
4
0
6
0
8
0
1
0
0
1
2
0
1
4
0
1
6
0
0
1
2
3
4
5
6
7
8
9
1
0
E
n
e
r
g
y

(
k
e
V
)
C o u n t s
0
2
0
4
0
6
0
8
0
1
0
0
1
2
0
1
4
0
1
6
0
0
1
2
3
4
5
6
7
8
9
1
0
E
n
e
r
g
y

(
k
e
V
)
C o u n t s
0
2
0
1
0
3
0
4
0
5
0
6
0
7
0
8
0
9
0
1
0
0
0
1
2
3
4
5
6
7
8
9
1
0
E
n
e
r
g
y

(
k
e
V
)
C o u n t s
0
5
0
1
0
0
1
5
0
2
0
0
2
5
0
3
0
0
3
5
0
4
0
0
0
1
2
3
4
5
6
7
8
9
1
0
E
n
e
r
g
y

(
k
e
V
)
C o u n t s
2
3
1
1
0
0

n
m
C
o
p
y
r
i
g
h
t

2
0
0
6

b
y

T
a
y
l
o
r

&

F
r
a
n
c
i
s
PT-30 resin. PT-30/PR-19-PS and PT-30/PR-24-PS were blended in 99.5/0.5 and
99/1 wt% (Fig. 16.33), and a total of four CNF blends were prepared.
PT-15 was blended with several clays (Nanomer
I.28E, Nanomer I.30E, PWG,

Cloisite 10A, and Cloisite 30B) using different mixing methods, and a total of 34
blends were recorded. Resulting samples were either opaque or translucent. PT-
15/30B (97.5/2.5) THF, PT-15/10A (97.5/2.5) THF, and PT-15/I.30E (97.5/2.5) THF
blends were prepared using THF as a carrier medium in these samples to facilitate
clay dispersion. When Cloisite 30B, Cloisite 10A, and Nanomer I.30E were compared
in higher magnification TEM micrographs, Cloisite 30B was dispersed slightly more
uniformly than the other two clays. Cloisite 30B was in a partial exfoliated state in
PT-15, as shown in Fig. 16.34. Small tactoids are present in this nanodispersion.
Figure 16.35 shows higher magnification TEM micrographs of PT-15/Cloisite 30B
FIGURE 16.33 TEM micrographs of PT-30/PR-24-PS (99/1) at high magnification showing
good dispersion of PR-24-PS CNF in the PT-30 cyanate ester. Scale bars = 500 nm (left) and
200 nm (right).
FIGURE 16.34 TEM micrographs of PT-15/Cloisite 30B (97.5/2.5) THF showing uniform
dispersion of Cloisite 30B in the PT-15 cyanate ester. Scale bar = 1 Pm.
(97.5/2.5) THF showing that Cloisite 30B clays are exfoliated in PT-15 cyanate ester.
Cloisite 30B is our preferred clay for the PT-15 resin.
PT-15 was blended with poly(phenyl silsesquioxane), PPSQ, in THF and trisilanol-
phenyl POSS in different wt%, and a total of 13 blends were recorded. Only PT-15/
trisilanol phenyl POSS in 99/1 and 97/3 were completely transparent. TEMs of PT-15/
SO1458 (97/3) and PT-15/SO1458 (95/5) blends were prepared. Figure 16.36 shows
molecular dispersion of SO1458 POSS in the PT-15/SO1458 (97/3) system. Where
particles are noted, one finds their Si content is higher than in the open resin matrix.
Significant Si is detected in the resin matrix in regions where no particles are
observed. Clearly the POSS is partially dissolved in the matrix and partially nano-
dispersed, resulting in molecular dispersion of SO1458 in PT-15. This is a good
candidate system for a conversion to CCC. SO1458 POSS is the preferred POSS
system for the PT-15 resin.
A summary of all the nanomodified carbon/carbon composite (NCCC) candi-
dates and CC139 (baseline commercial carbon/carbon composite prepared with
134A phenolic resin) with their weight loss percentage, density, and ranking based
on weight loss data is shown in Table 16.8. Samples were designated of 1 through
4 to indicate room temperature, 700C, 1200, and retained conditions, respectively.
An air leakage was detected in our first set of experiments, and as a result the
134A/CLO/3-1A, 134A/CLO/3-2A, 134A/CLO/3-3A, 134A/POSS/3-1A,
134A/POSS/3-2A, 134A/POSS/3-3A, PT15/ POSS/5/-1B, PT15/POSS/5/-2A,
PT15/POSS/5/-3A, PT30/PR24PS/1-1A, and PT30/PR24PS/1-2A specimens were
overexposed to unrealistic thermo-oxidative conditions. They are denoted with aster-
isks in Table 16.8. The 134A/CLO/3-4B and 134A/POSS/3-4D specimens were
retested at 1200qF using the fourth spared panels of the 134A/CLO/3 and
134A/POSS/5 candidates designated. The PT15/POSS/5-4 spared specimen was lost,
and as a result the PT15/POSS/5 candidate was unable to be retested for the 1200qF
conditions of this project. Regrettably the behavior of molecularly dispersing POSS
into PT-15 (Table 16.8, entries 9 to 12) could not be assessed for improved thermo-
oxidative stability due to overexposure or loss.
FIGURE 16.35 Higher magnification TEM micrographs of PT-15/Cloisite 30B (97.5/2.5)
THF showing Cloisite 30B clays are exfoliated in PT-15 cyanate ester.
Based on the thermo-oxidative analyses, the PT30/CLO/5 candidate is the most
thermo-oxidative resistant material; weight losses were 5.2%, 5.1%, and 3.7% at
room temperature (RT), 700qF, and 1200qF, respectively. PT30 ranked as first or
best. The PT15/CLO/5 candidate was the second best thermo-oxidative resistant
material and weight losses were 6.5%, 5.1%, and 3.7% at RT, 700qF, and 1200qF
conditions, respectively. The PT30/PR24PS/1 NCCC showed a low 6.8% weight
loss and was the third best thermo-oxidative resistant material NCCC candidate. The
standard CC139 was fourth with weight losses of 8.9%, 9.0%, and 13.6% for RT,
700qF, and 1200qF, respectively. The 134A/CLO/3-4B and 134A/POSS/3 had high
weight losses of 30.5% and 23.3%, respectively.
Weight losses of different NCCCs were compared, as shown in Fig. 16.37. It is
evident that the cyanate ester resin-based PT30/PR24PS/1, PT30/CLO/5, and
PT15CLO/5 specimens had better thermo-oxidative resistance than the phenolic
resin-based CC139 standard CCCs. The phenolic resin-based 134A/CLO/3 and
134A/POSS/3 specimens exhibited worse thermo-oxidative resistant characteristics
FIGURE 16.36 Progressive magnification TEM micrographs of PT-15/SO1458 POSS (97/3)
showing molecular dispersion of SO1458 POSS in PT-15 cyanate ester. Si from POSS
molecules is dispersed in the resin matrix. The dark particle has the same Si/O/C composition
as the resin matrix. Significant Si signal is detected from the matrix. The signal of Cu is from
the grids.
Diffuse areas in matrix
0 1 2 3 4 5 6 7 8 9 10
0
10
20
30
40
50
60
70
80
C
o
u
n
t
s
0 1 2 3 4 5 6 7 8 9 10
Energy (keV) Energy (keV)
0
10
20
30
40
50
60
70
80
C
o
u
n
t
s
C
Cu
Cu
Cu
O
#255 Matrix
Si
C
Cu
Cu
Cu
O
#254 Particle
Si
(The signal of Cu is from the grids) Significant Si signal is
detected from the matrix.
than the CC139 standard. From our TEM analyses we concluded that the nanopar-
ticles were dispersed very well in the PT-30 and PT-15 cyanate resin systems and
poorly in the 134A phenolic resin system. Figure 16.38 shows a comparison of
the densities of the different NCCCs. It is also evident that PT15/CLO/5-3B,C
had the highest density (1.684 g/cc), followed by PT30/CLO/5-2B,C and
PT30/PR24PS/1-3B,C (1.656 g/cc), compared with the CC139 specimens (1.627
to 1.638 g/cc), while the densities of 134A/POSS/3-4D and 134A/CLO/3-4B were
1.603 g/cc and 1.585 g/cc, respectively. Figure 16.39 compares densities versus
weight losses of the different NCCCs. A trend of higher density NCCCs exhibiting
TABLE 16.8
Summary of Thermo-Oxidative Data and Ranking
P/N
Initial
wt. (g)
TGA
recorded
wt. (g)
Weight
Loss
(%)
Sample
Density
(g/cc) Rank
134A/CLO/3-1A
*
2.053 1.728 15.8 1.619
134A/CLO/3-2A
*
2.145 1.772 17.4 1.581
134A/CLO/3-3A
*
2.056 1.152 44.0 1.567
134A/CLO/3-4B 2.036 1.414 30.5 1.585 6
134A/POSS/3-1A
*
1.958 1.441 26.4 1.616
134A/POSS/3-2A
*
1.989 1.216 38.9 1.575
134A/POSS/3-3A
*
1.933 1.666 13.8 1.611
134A/POSS/3-4D 2.055 1.573 23.2 1.603 5
PT15/POSS/5-1B
*
2.469 2.178 11.8 1.650
PT15/POSS/5-2A
*
2.462 2.146 12.9 1.595
PT15/POSS/5-3A
*
2.577 1.782 30.8 1.651
PT15/POSS/5-4 Lost
PT15/CLO/5-1 2.52 2.36 6.5 1.630 2
PT15/CLO/5-2B 2.36 2.28 3.4 1.643 2
PT15/CLO/5-2C 2.40 2.24 6.8 1.643 2
PT15/CLO/5-3B 2.34 2.26 3.5 1.684 2
PT15/CLO/5-3C 2.38 2.29 3.8 1.684 2
PT30/PR24PS/1-1A
*
2.586 1.244 51.9 1.653
PT30/PR24PS/1-2A
*
2.574 1.406 45.4 1.618
PT30/PR24PS/1-3B,C 2.573 2.399 6.8 1.656 3
PT30/CLO/5-1B,C 2.522 2.358 5.2 1.653 1
PT30/CLO/5-2B,C 2.383 2.261 5.1 1.656 1
PT30/CLO/5-3B,C 2.363 2.265 3.7 1.618 1
CC139-1A 2.673 2.436 8.9 1.627 4
CC139-2A 2.506 2.881 9.0 1.638 4
CC139-3A 2.536 2.191 13.6 1.630 4
*
Air leakages were detected in these experiments. room temperature, 700qF, and 1200qF
specimens were designated as 1, 2, and 3, respectively. Asterisks indicate overexposed
specimens.
lower weight loss while lower density NCCCs show higher weight loss is proposed.
Lower weight loss for NCCCs from cyanate ester resin systems is expected due to
few or no volatiles from cured cyanate esters as well as proposed improved thermo-
oxidative stability attributable to the presence of nanophases in the NCCC cyanate
ester materials.
FIGURE 16.37 Comparison of weight losses of different NCCCs.
FIGURE 16.38 Comparison of densities of different NCCCs.
1
3
4
A
/
C
L
O
/
3
-
4
B
1
3
4
A
/
P
O
S
S
/
3
-
4
B
P
T
3
0
/
P
R
2
4
P
S
/
1
-
3
B
,
C
P
T
3
0
/
C
L
O
/
5
-
1
B
,
C
P
T
3
0
/
C
L
O
/
5
-
2
B
,
C
P
T
3
0
/
C
L
O
/
5
-
3
B
,
C
P
T
1
5
/
C
L
O
/
5
-
2
B
,
C
P
T
1
5
/
C
L
O
/
5
-
3
B
,
C
C
C
1
3
9
-
1
A
C
C
1
3
9
-
2
A
C
C
1
3
9
-
3
A
P
T
1
5
/
C
L
O
/
5
-
1
0
5
10
15
20
25
30
35
W
e
i
g
h
t

l
o
s
s

(
%
)
1.52
1.54
1.56
1.58
1.6
1.62
1.64
1.66
1.68
1.7
1
3
4
A
/
C
L
O
/
3
-
4
B
1
3
4
A
/
P
O
S
S
/
3
-
4
D
P
T
3
0
/
P
R
2
4
P
S
/
1
-
3
B
,
C
P
T
3
0
/
C
L
O
/
5
-
1
B
,
C
P
T
3
0
/
C
L
O
/
5
-
2
B
,
C
P
T
3
0
/
C
L
O
/
5
-
3
B
,
C
P
T
1
5
/
C
L
O
/
5
-
1
P
T
1
5
/
C
L
O
/
5
-
2
B
,
C
P
T
1
5
/
C
L
O
/
5
-
3
B
,
C
C
C
1
3
9
-
1
A
C
C
1
3
9
-
2
A
C
C
1
3
9
-
3
A
D
e
n
s
i
t
y

(
g
/
c
c
)
16.4 SUMMARY AND CONCLUSIONS
We have been developing polymer nanocomposites to enhance materials properties
for high-temperature applications. The following conclusions are presented:
1. The feasibility of using polymer nanocomposite for high-temperature
applications was clearly demonstrated.
2. MMT organoclays, CNFs, and POSS can be easily incorporated into var-
ious polymers using conventional processing and manufacturing tech-
niques to form nanocomposites.
3. The degree of dispersion of nanoparticles in the polymer matrix is essen-
tial to achieve the desired properties enhancement.
4. SEM and TEM analyses techniques have been demonstrated to be effective
and efficient screening tools to determine nanodispersion in the polymer
matrix.
PROBLEMS
1. What are the shortcomings of using wide-angle x-ray diffraction (WAXD)
to determine the degree of dispersion in polymerclay nanocomposites?
How can one compensate for the shortcomings of this technique?
2. What is the disadvantage of using transmission electron microscopy
(TEM) to determine the degree of nanoparticle dispersion in polymer
nanocomposites? How can one compensate for this technique?
3. Discuss other types of techniques that one can use to study the morphology
of these polymer nanocomposites.
FIGURE 16.39 Comparison of densities and weight losses of different NCCCs.
0
5
10
15
20
25
30
35
1.58 1.6 1.62 1.64 1.66 1.68
Density (g/cc)
W
e
i
g
h
t

l
o
s
s

(
%
)
4. What other types of nanoparticles can be used to enhance the thermal and
mechanical properties of polymer nanostructured materials? Explain and
give examples.
5. Discuss the similarities and differences of different types of nanoparticles
and give examples.
6. Discuss other thermal applications of polymer nanostructured materials
in the literature.
7. How can one enhance the interfacial interaction and compatibility with
different nanoparticles in a specific polymer matrix?
Readers should read the references in this chapter and current nanocomposite liter-
ature to answer the preceding questions.
ACKNOWLEDGMENTS
The authors would like to thank Dr. Charles Y-C. Lee of the Air Force Office of
Scientific Research for sponsoring several of our research activities through the
AFOSR STTR programs. Support from Dr. Shawn Phillips of the Air Force Research
Laboratory and Edwards Air Force Base and W. Casey West of StanChem is appre-
ciated. The authors also would like to express their appreciation to numerous col-
leagues who have contributed to our nanomaterials research.
REFERENCES
1. Krishnamoorti, R. and Vaia, RA., Ed., Polymer Nanocomposites: Synthesis, Charac-
terization, and Modeling, Symposium Series 804, American Chemical Society, Wash-
ington, D.C., 2001.
2. Cloisite 30B technical data sheet, Southern Clay Products, Gonzales, TX.
3. Koo, J.H., Stretz, H. et al., Phenolic-clay nanocomposite ablatives for rocket propulsion
systems, Proceedings of International Symposium, SAMPE, Covina, CA, 1085, 2002.
4. Koo, J.H., Stretz, H. et al., Nanocomposite rocket ablative materials: processing,
characterization, and performance, Proceedings of International Symposium,
SAMPE, Covina, CA, 1156, 2003.
5. Koo, J.H., Chow, W.K. et al., Flammability properties of polymer nanostructured
materials, Proceedings of International Symposium, SAMPE, Covina, CA, 954, 2003.
6. Koo, J.H., Stretz, H. et al., Nanocomposites rocket ablative materials: subscale ablation
test, Proceedings of International Symposium, SAMPE, Covina, CA, 2004.
7. Koo, J.H., Pittman, C. et al., Nanomodified carbon/carbon composites for intermediate
temperature: processing and characterization, Proceedings of 35th International
SAMPE Technical Conference, SAMPE, Covina, CA, 2003.
8. Koo, J.H., Pilato, L. et al., Nanomodified carbon/carbon composites for intermediate
temperature, AFOSR STTR Phase I Final Report, submitted to Air Force Office of
Scientific Research, Arlington, VA, Jan. 2004.
9. Koo, J.H., Pilato, L. et al., Thermo-oxidative studies and mechanical properties of
nanomodified carbon/carbon composites, Proceedings of International Symposium,
SAMPE, Covina, CA, 2004.
10. Blanski, R., Koo, J.H. et al., Polymer nanostructured materials for solid rocket motor
insulation: ablation performance, Proceedings of 52nd JANNAF Propulsion Meeting,
CPIA, Columbia, MD, 2004.
11. Koo, J.H., Marchant, D. et al., Polymer nanostructured materials for solid rocket motor
insulationProcessing, microstructure, and mechanical properties, Proceedings of 52nd
JANNAF Propulsion Meeting, CPIA, Columbia, MD, 2004.
12. Koo, J.H., Pilato, L.A. et al., Nanocomposites for carbon fiber-reinforced polymer
matrix composites, AFOSR STTR Final Report, Arlington, VA, Oct. 2004.
13. Koo, J.H., Pilato, L.A., Wissler, G., Ervin, M., and Firestone, K., Innovative selective
laser sintering rapid manufacturing technique using nanotechnology, unpublished data.
14. Tibbetts, G.G. and McHugh, J.J., Mechanical properties of vapor-grown carbon fiber
composites with thermoplastic materials. J. Mater. Res., 14, 28712880, 1999.
15. Lake, M.L. and Ting, J.-M., Vapor-grown carbon fiber composites in carbon materials
for advanced technologies., in Carbon Materials for Advanced Technologies, Burchell,
T.D., Ed., Pergamon, Oxford, UK, 1999.
16. Lake, M.L., Low-Cost Carbon Fibers for Industrial Applications. Applied Sciences,
Inc., Cedarville, OH, 11/2002.
17. Maruyama, B. and Alam, K., Carbon nanotubes and nanofibers in composite mate-
rials. SAMPE J., 38, 3, 59, 2002.
18. Glasgow, D.G., Tibbetts, G.G., Matuszewski, M.J., Walters, K.R., and Lake, M.L.,
Surface treatment of carbon nanofibers for improved composite mechanical proper-
ties, Proceedings of International Symposium, SAMPE, Covina, CA, 2004.
19. Lee, A., Proceedings of POSS
Nanotechnology Conference, Huntington Beach, CA,

Sept. 2527, 2002.
20. Hu, X., Koo, J.H. et al., Flammability studies of water-borne fire retardant nanocom-
posites coating on wood, Proceedings of Polymer Nanocomposites Symposium, Amer-
ican Chemical Society Southwest Regional Meeting, San Antonio, TX, Oct. 1720,
2001.
21. Patton, R.D., Pittman, C.U., Wang, L., Hill, J.R., and Day, A., Ablation, mechanical
and thermal conductivity properties of vapor growth carbon fiber/phenolic matrix
composites, Composites Part A, 33, 243251, 2002.
17
Manufacturing,
Mechanical
Characterization, and
Modeling of a Pultruded
Thermoplastic
Nanocomposite
Samit Roy, Kalivarathan Vengadassalam,
Farzana Hussain, and Hongbing Lu
CONTENTS
17.1 Introduction
17.1.1 Nanoclays
17.1.2 Montmorillonite
17.1.3 Dispersion of Montmorillonite Clay Platelets
in Polypropylene
17.1.4 Compression Strength of Nanocomposites
17.2.1 Materials
17.2.2 Processing of FiberReinforced Nanocomposites
17.2.2.1 Thermoplastic Pultrusion Process
17.2.2.2 Process Parameter Control
17.2.3 Mechanical Test Methodology
17.3 Nanocomposite Morphology
17.3.1 Transmission Electron Microscopy (TEM)
17.3.2 Scanning Electron Microscopy (SEM)
17.4 Results and Discussion of Test Data
17.4.1 Transmission Electron Microscopy
17.5 Mechanical Properties Characterization
17.5.1 Uniaxial Compression Tests
17.5.2 Inelastic Kinking Analysis for Pure Compression
Loading
17.5.3 Calculation of Fiber Volume Fraction, Composite Shear
Modulus, and I/J*
y
Ratio
17.5.4 Discussion of Results
Problems
Acknowledgments
References
17.1 INTRODUCTION
The use of polymer matrix composites (PMCs) in aerospace and civil engineering
as well as in sports and leisure applications is rapidly increasing. PMCs have found
a wide range of applications in structural components in which the substitution of
PMCs for metals has substantially improved performance and reliability. The high
tensile strength of PMCs is mainly derived from the high strength of the carbon or
glass fibers embedded in the matrix. Fibers typically have high strength in tension.
However, their compressive strength is generally much lower due to the fact that
under compression the fibers tend to fail through buckling well before compressive
fracture occurs. Also, fiber misalignment and presence of voids during manufac-
turing processes contribute to a further reduction in compressive strength. In fact,
the overall compressive strength of a PMC is only about 50% of the tensile strength.
The strength of the surrounding polymer matrix plays a key role in characterizing
the critical buckling loads of the fibers by constraining the fibers from buckling.
Consequently high-strength polymer resins such as PEEK and PPS although
very expensive are typically used for applications requiring high compressive
strength. Low-cost commodity resins such as polypropylene (PP) suffer primarily
due to low compressive strength. Because the critical buckling stress in the fiber
is directly related to the stiffness and yield strength of the surrounding matrix
material, any increase in these matrix properties would directly result in an improve-
ment in the compressive strength of the composite. In this context, it has been
observed that the addition of small amounts of nanostructured montmorillonite
(MMT) clay (~5 wt%) significantly improves the stiffness, strength, gas barrier,
and fire resistance properties of most thermoplastic resins, with only small effect
on flexibility. These nanoclay fillers are only slightly more expensive than glass (a
few dollars per pound), yet generally far less expensive than carbon fibers
(~$100/lb) or carbon nanotubes (~$50,000/lb). Additionally the small amounts of
nanofillers required to enhance properties enables these materials to compete more
effectively with traditional glass-fiber reinforcements.
17.1.1 NANOCLAYS
A nanoclay is a mineral with a high aspect ratio and with at least one dimension of
the particle typically the thickness in the nanometer range. Reinforcements in
the nanometer size range closely approach the molecular size of the polymer,
enabling an intimate encounter between the two materials. When properly modified,
the filler particles and polymer interact to create constrained regions at the particle
surface. This immobilizes a portion of the polymer chain, creating a reinforcement
effect. Purity and cation exchange capacity are two characteristics critical to the
success of nanoclays as polymer reinforcing agents.
Because a nanofiller contains so many individual particles in such a small amount
of material, it requires very low loading to obtain a high concentration of constrained
areas within the polymer. For example, 5% loading by weight of nanoclay leads to a
reinforcing effect equal to about 12 to 15% glass fiber. The nanofiller also creates
a tortuous path for the penetration of gaseous vapors and liquids into the polymer,
which leads to better chemical and moisture resistance.
17.1.2 MONTMORILLONITE
Montmorillonite (MMT) has the widest acceptability for use in polymers. It is a
type of smectite clay that can absorb water, and it has a layered structure, with
aluminum octahedron sandwiched between two layers of silicon tetrahedron. Each
layered sheet is slightly less than 1 nm thin (10 ), with surface dimensions extend-
ing to about 1 Pm or 1000 nm. The aspect ratio is about 1000:1 and the surface area
is in the range of 750 m
2
/g.
Depending on the degree of polymer penetration into the silicate tactoid, two
idealized nanocomposite structures are feasible: intercalated and exfoliated. The
process of opening the spaces between the clay platelets, which are known as
galleries, is called intercalation. Intercalation increases the separation distance
between the nanoclay platelets within a clay tactoid and is brought about by surfac-
tant molecules. Without this separation, the polymer molecules will not be able to
penetrate into the galleries between the platelet layers. In the intercalated systems,
the multilayer structure of the silicates is retained, with alternating polymer/silicate
layers and a repeat spacing larger than that of the original clay.
In exfoliated nanocomposites, the primary particles of the organically modified
clay are delaminated into individual nanometer-thick silicate platelets, which are
dispersed uniformly in the polymer matrix due to extensive polymer penetration and
intercalation. In the exfoliated form, nanofillers have very small flexible platelet-
type structures. As mentioned earlier, the thickness of the platelet is in the nanometer
range, while the width and length are between 0.1 and 2 Pm. Because of this, a
single gram of exfoliated nanoclay will contain over a million individual particles.
17.1.3 DISPERSION OF MONTMORILLONITE CLAY PLATELETS
IN POLYPROPYLENE
For nanoclay to improve mechanical strength of a polymer at a low clay loading
level, the clay must be exfoliated into platelets having the thickness of one unit cell
of the crystal structure, about 1 nm. Dispersion of montmorillonite (MMT) into
polypropylene (PP) is especially challenging because the ionic surface of the clay
repels the aliphatic hydrocarbon polymer. Dispersion of MMT into PP in an extruder
has been accomplished with (1) surfactant modified clay, (2) PP modified with maleic
anhydride (PP-MA), and (3) a combination of both modified clay and PP-MA. The
most common surfactant for making the clay organophilic is octadecylamine (ODA).
Other amines and ammonium ions such as trialkylimidazolium ions, polyamine
dendrimers, and amine-terminated polyethylene and polypropylene are also known
to aid dispersion of clays. Composites containing both glass fibers and clays modified
with cationic surfactants also appear promising. Favorable enthalpic interaction
between the clay and the PP, which is not achieved by use of ODA-modified clay
and unmodified PP, would aid dispersion of exfoliated clay.
17.1.4 COMPRESSION STRENGTH OF NANOCOMPOSITES
Earlier work by Rosen,
1
which assumed extensional (transverse) and shear microbuck-
ling of the fibers embedded in an elastic matrix, consistently overpredicted the com-
pressive strength when compared with experimentally measured values. Argon
2
was
among the first to recognize that fiber-reinforced composites made by normal manu-
facturing processes, including pultrusion, have regions of fiber misalignment. Argon,
2
and later Budiansky,
3,4
showed that fiber misalignments present in fiber-reinforced
composites could lead to yielding of the polymer matrix in shear. The yielding of the
matrix would, in turn, result in loss of matrix stiffness that could eventually trigger
fiber microbuckling with resulting kink band formation leading to final failure.
4
Consequently, it was recognized that the dominant compressive failure mode in
aligned-fiber polymer matrix composites is localized compressive buckling or kinking.
Argon
2
proposed a simple yet elegant formula for composite compressive strength
given by V
C
V
Y
/I
0
where V
Y
is the matrix yield strength in shear, and I
o
is the initial
fiber misalignment angle with respect to load direction, sometimes referred to as fiber
misorientation. Further it is likely that any increase in composite longitudinal shear
stiffness (G
12c
) due to the presence of nanoclay particles would result in an increase
in the compressive strength of the composite, as indicated by a modified form of the
Argon formula proposed by Budiansky that includes both elastic and plastic buckling
behavior as limiting cases,
46
(17.1)
where J
Y
V
Y
/G
12c
is the composite yield strain in longitudinal shear. The failure
mechanism described in Equation 17.1 is still a shear mode of fiber buckling, but unlike
Rosens elastic analysis, the local nature of the imperfections, coupled with (perfectly)
plastic yielding of the matrix in shear, results in local microbuckling and kink band
formation, as shown in Fig. 17.1. However, it should be noted that Equation 17.1 assumes
zero kink band angle and an elasticperfectly plastic matrix material.
From Equation 17.1, it quickly becomes apparent that improvements in the com-
pressive strength of the composite may be achieved by (1) improving the yield strength
and stiffness of the surrounding matrix in shear and (2) reducing fiber misalignment
in the composite through optimization of manufacturing process variables, such as
pull speed, preformer temperature, nanoparticle alignment, and dispersion. It is there-
fore the aim to accomplish this objective by concurrently increasing matrix yield
strength and matrix shear modulus, as well as by decreasing fiber misalignment,
thereby making use of synergies that may exist between these effects. There exist very
c
c
Y
G
+
j
(
,
\
,
(
12
0
1
few published data on the compressive strength of pultruded thermoplastic composites.
Fibers in pultruded material are in general well aligned. However, the composite
compressive strength is very sensitive to even slight misalignment in fiber orientation.
5,6
Typically a thermoplastic resin matrix offers better structural performance than
thermosetting resin matrix. Thermoplastics exhibit superior fracture toughness
7
and
impact resistance, and have better postprocess formability and improved reparability.
With improved manufacturing techniques, such as ultrasonically activated consolida-
tion dies, it is now feasible for thermoplastic pultrusion to yield higher production
rates than thermoset composites. Recyclability of thermoplastics is another attractive
feature. In this chapter, the effect of incorporating nanoclay particles on mechanical
properties such as compressive strength and compressive modulus of pultruded E-
glass/PP unidirectional composites will be presented and discussed.
17.2.1 MATERIALS
Marlex HGL-350 polypropylene homopolymer (antistatic) was used as neat resin
(density 902 kg/m
3
and melt flow rate 35 g/10 min). Material systems evaluated in the
experimental studies were unidirectional E-glass fibers melt-impregnated with polypro-
pylene with and without nanoclay particles using a single-screw extruder. To pultrude
unidirectional composites shapes using either E glass/PP or nanoclay/E glass/PP,
prepreg tapes were fed to the pultrusion machine. Subsequently 2.54-cm wide and
1.27-cm thick composite beams for different nanoclay content (0%, 1%, 2%, 3%, 4%,
5%, and 10% by weight) were pultruded. The pultruded samples were machined
according to ASTM D695 specifications to obtain compression test specimens. The
clay used was organically modified montmorillonite, 1.28 E supplied by Nanocor.
FIGURE 17.1 Compression failure of pultruded specimen showing kink band.
17.2.2 PROCESSING OF FIBERREINFORCED NANOCOMPOSITES
17.2.2.1 Thermoplastic Pultrusion Process
The use of thermoplastics as matrix materials for polymer composites has gained
popularity recently due to their high-performance properties compared with ther-
moset resins.
8
Pultrusion is among the most rapid and efficient processes to fabricate
composites when low cost or high volume is required.
9,10
The versatility of this
process enables the production of profiles with many different sectional shapes, in
what has largely contributed to the rapid increase of pultruded composites in the
construction industry. It remains one of the most efficient and inexpensive methods
for producing continuous fiber composites. In its simplest form, continuous rein-
forcing fibers (or tapes) are coated with a resin, and then shaped and cured by pulling
the mass through a heated die. The process consists of the following general steps:
(1) preimpregnated reinforcement feeding (the preimpregnation of fibers is usually
done in a separate step using an extruder); (2) die shaping/curing; (3) pulling; and
(4) cutting. The preimpregnated reinforcing materials (continuous fibers, mats, etc.)
are guided through carding plates and other style guides to assemble them in the
right place and order. Next the fiber and resin mass flows through a heated preformer
die that causes the thermoplastic resin to melt before it enters the consolidation die.
The operating parameters (temperature and pull speed) are adjusted so that the
composite is fully consolidated as it leaves the die. Finally a mechanical means for
pulling is provided, e.g., a belt puller. A system for cutting the pultruded composite
as it is continuously pulled is also provided. In the case of the novel process utilized
by the pultrusion machine at Oklahoma State University (OSU), the prepreg is guided
through a heated preformer and then through an ultrasonically activated final con-
solidation die. A schematic of the pultrusion machine with its ultrasonic die assembly
is shown in Fig. 17.2. The ultrasonic vibrations directed into the preheated prepreg
by means of a waveguide help stimulate the flow of resin in the prepreg. Pressure
is also applied to the prepreg within the die to consolidate the prepreg to the chosen
profile. The ultrasonic vibration at low power input (maximum power 50 watts) does
result in some incidental heating in the prepreg, but the increased flow rates of the
resin indicate a reduction in viscosity that is far greater than expected in relation to
FIGURE 17.2 Schematic of pultrusion machine with ultrasonic consolidation die.
the temperature increase. The net effect of the ultrasonic die is to reduce effective
resin viscosity as well as friction at the compositedie interface, thereby significantly
reducing void content and pull-force requirement. Pultruded E glass/PP samples that
are 2.54 cm wide and 1.27 cm thick are shown in Fig. 17.3 for 0%, 1%, 2%, and
3% nanoclay loading by weight. The change in color of the samples due to the
presence of clay is noticeable.
17.2.2.2 Process Parameter Control
The most critical processing parameters to be controlled are speed and preformer
temperature profile. For any fixed die length the process speed will determine the
residence time of material. Speed also depends on the sectional thickness of the part.
In this research, E glass/PP composites were pultruded successfully at speeds up to
5 mm/s. Initially pulling speed was maintained as low as 0.5 mm/s to ensure correct
consolidation of the thermoplastic composite. The preformer die consisted of three
temperature zones. Temperature set points as measured by the thermocouples at these
three zones in the preformer die were 220qC, 300qC, and 325qC, respectively. The
preformer die temperature profile determined the rate of heating of material and
influenced the resin viscosity and degree of cure attained during residence in the
preformer. Subsequently the composite passed through the final consolidation die,
where ultrasonic vibrations directed into the preheated prepreg by means of a
waveguide helped stimulate the flow of resin in the prepreg. Pressure was applied
to the prepreg within the die to consolidate the prepreg to the chosen profile. Finally
the pultruded laminates were sectioned using a precision cutting saw that traveled at
the same speed as the part being pultruded and examined in an optical microscope
in order to determine the size and distribution of voids and unwetted fibers. Table
17.1 summarizes the pultrusion processing condition.
FIGURE 17.3 Pultruded E glass/PP samples with different nanoclay loadings.
17.2.3 MECHANICAL TEST METHODOLOGY
Two separate sets of compression tests were performed. In the first set, ASTMs
D695 procedure was used to determine the compressive properties of neat PP and
nanoclay-modified PP resin samples using an MTS (22.24 kN range) testing machine
with two parallel platens. Load was measured by load cells attached to the tops of
TABLE 17.1
Pultrusion Processing Variables
Prepreg
Precursor
No. of
Tapes
Preformer Temperature
Profile (qC)
Die
Pressure
(kPa)
Pulling
Speed
(mm/s)
Ultrasound
Energy
(Watts)
Zone 1 Zone 2 Zone 3
PP/E glass 38
PP/E glass
1% clay
37
PP/E glass
2% clay
37
220 300 325 3570 0.55.0 2225
PP/E glass
3% clay
37
PP/E glass
4% clay
36
PP/E glass
5% clay
36
PP/E glass
10% clay
36
FIGURE 17.4 Cylindrical polypropylene neat resin specimen after compression testing.
the platens. The tests were carried out using compression-molded cylindrical test
specimens that were 12.7 mm in diameter and 25.4 mm in length, as shown in Fig.
17.4. From applied load and displacement data, stress versus strain curves were
plotted using Instron fast track data acquisition software.
In the second set of tests, compressive strength and compressive modulus of
pultruded E glass/PP with different loadings of nanoclay particles were determined
by performing compressive tests using a standard 245-kN Servo hydraulic MTS
machine in displacement mode. Again the ASTM D-695 test method was used to
determine the compressive strength and modulus of elasticity of the E glass/PP
nanocomposites. A test specimen after failure is shown in Fig. 17.5, with a close-
up view of the resulting kink band shown in Fig. 17.1. At least four replicate test
specimens were used for each nanoclay loading.
17.3 NANOCOMPOSITE MORPHOLOGY
17.3.1 TRANSMISSION ELECTRON MICROSCOPY (TEM)
For thin foil preparation, cylindrical specimens of 3-mm diameter were cut by
an ultrasonic disc cutter. This procedure was performed for compression-molded
resin specimens as well as for pultruded E glass/PP nanocomposite samples.
These cylinders were then sliced in a diamond saw to 0.5 mm or smaller and
then ground gently with a 600-grade silicon carbide paper to a thickness of about
200 m and dimpled by mechanical polishing. These dimpled discs were then
ion milled and thinned to a thickness of less than 100 nm to allow electron beam
transmission. The thin foils so prepared were examined with a 200-KeV TEM
(Hitachi H 8000) equipped with EDS facilities to determine elemental composi-
tion of the specimens.
FIGURE 17.5 Pultruded compression test specimen (10% nanoclay loading) after failure.
17.3.2 SCANNING ELECTRON MICROSCOPY (SEM)
Microstructure and fracture surfaces of the composites were evaluated by scanning
electron microscopy (JOEL) JXM 6400, at an accelerating voltage of 20 kV. A
Balzer MED 010 sputtering system was used with an Au-Pd sputter coater.
17.4 RESULTS AND DISCUSSION OF TEST DATA
17.4.1 TRANSMISSION ELECTRON MICROSCOPY
Transmission electron microscopy (TEM) was applied to characterize the hierar-
chical morphology present in the nanocomposites. Figure 17.6(a) shows a bright-
field TEM micrograph of polypropylene with 3% clay loading in a dogbone sample.
Figure 17.6(b) shows the selected area diffraction pattern for 3% polypropylene-
based nanocomposites. The diffraction pattern shows several reflections with
varying degrees of intensity. Figure 17.6(a) and Figure (b) reveal the presence of
aggregates, but there are regions where completely delaminated sheets are dis-
persed individually, showing the dark lines of 10- thickness. It appears that the
clay exists as expanded aggregates made up of 2 to 10 platelets. The sizes of these
FIGURE 17.6 (a) Bright-field TEM micrograph of polypropylene with 3% clay loading in
a dogbone sample. Note the irregular lamellar morphology of the polymer microstructure. A
small nanosize clay particle is present in the picture. (b) Selected area diffraction pattern.
Scale bar = 225 nm.
aggregates and the numbers of platelets in them increase with the percentage of
clay within the PP matrix. Indeed the TEM pictures show all possible platelet
morphologies, namely exfoliated, intercalated, and stacked structures within the
samples. There appears to be reasonably good compatibility of the surface-treated
clay with the polypropylene resin.
Figure 17.7(a) shows a typical arrangement of a group of such glass fibers present
in a thin film of E glass/polypropylene with 3% nanoclay loading. Nanosized clay
particles present in the composites were also found randomly distributed in the
polypropylene within the individual spacing of E glass. Figure 17.7(b) a bright-field
micrograph indicating the existence of a nanosized clay particle embedded in the
matrix of polypropylene and in contact with the E glass fibers. This provides evidence
of particlefiber interaction in which the clay particles are actually in contact with
the fiber wall. The orientation of the clay with respect to the fiber surface is unclear.
A close-up view in Fig. 17.8 reveals the existence of a nanoclay agglomerate (A), as
well as intercalated particles (B) and an exfoliated particle (C).
FIGURE 17.7 (a) Bright-field TEM micrograph of compression molded E glass/polypropylene
with 3% clay loading sample. The micrograph shows a few relatively large particles on the
surface of the fiber, positions designated by arrows. (b) TEM bright-field micrograph showing
how polymer is entrapped between two fibers. Positions of nanoclay particles (3% nanoclay)
within polymer are indicated by arrows.
17.5 MECHANICAL PROPERTIES CHARACTERIZATION
17.5.1 UNIAXIAL COMPRESSION TESTS
Compression strength data for neat polypropylene (PP) resin for different nanoclay
loading are tabulated in Table 17.2. Figure 17.9 is a bar chart showing the progressive
improvement in the normalized compressive strength of nanoclay-reinforced PP for
different nanoclay content. A 134% increase in compressive strength was observed
for PP resin with 10% clay loading.
FIGURE 17.8 Bright-field TEM micrograph of polypropylene with 3% clay loading fiber-
reinforced composite sample. Fine structure of polymer is evident in the micrograph. Positions
denoted by A, B, and C with arrows indicate groups of particles, a couple particles, and single
particle, respectively. Scale bar = 375 nm.
TABLE 17.2
Compression Test Data for Neat Polypropylene Resin and Nanoclay-
Modified Resin
Sample
Clay
(wt%)
Avg.
Compressive
Strength (MPa)
% Increase
Compressive
Strength
Avg.
Standard
Deviation
PP 0.0 6.51 0.045
PP 1.0 8.29 26 0.29
PP 2.0 9.42 42 0.22
PP 3.0 10.33 57 0.23
PP 4.0 11.01 67 0.32
PP 5.0 12.77 94 0.48
PP 10.0 15.37 134 0.89
Stress versus strain curves for nanoclay-reinforced E glass/PP composite are
shown in Fig. 17.10 for different nanoclay loadings, and the compressive strength
data are summarized in column 3 of Table 17.3. Because our primary objective was
to obtain strength data, the data shown in Fig. 17.10 were not corrected for machine
compliance.
FIGURE 17.9 Normalized compressive strength versus clay content for neat polypropylene.
FIGURE 17.10 Comparison of stress versus strain curves from compression test for E
glass/PP pultruded nanocomposites with different nanoclay loadings.
134%
94%
67%
57%
42%
26%
0
0.5
1
1.5
2
2.5
3
Wt % of clay loading
N
o
r
m
a
l
i
z
e
d

c
o
m
p
r
e
s
s
i
v
e

s
t
r
e
n
g
t
h
0% 1%
2%
3% 4% 5% 10%
0
50
100
150
200
250
0.002 0.012 0.022 0.032 0.042 0.052 0.062
Strain (mm/mm)
A
v
g

c
o
m
p
r
e
s
s
i
v
e

s
t
r
e
s
s

(
M
P
a
)
PP E-glass-0% clay
PP E-glass-1% clay
PP E-glass-2% clay
PP E-glass-4% clay
PP E-glass 5% clay
PP E-glass-10% clay
Figure 17.11 shows the progressive improvement in the normalized compressive
strength of pultruded E glass/PP composites with different nanoclay content. Each
bar represents the average of four test specimens, with a standard deviation of
roughly 5%. A 122% improvement in strength for 10% clay content was observed
compared with baseline specimens with zero nanoclay content. A similar dramatic
enhancement was measured for normalized compressive modulus (stiffness), as
shown in Fig. 17.12, where 10% clay reinforcement led to a 110% improvement in
TABLE 17.3
Shear Modulus, Compressive Strength, Strain at Yield and Shear Strain in
the Kink Band for Pultruded E Glass/PP/MMT
% Clay
G
Composite
(GPa)
V
Critical
(GPa)
V
Critical
/
G
Composite
J
y at
0.2176 J
c
J
c
/J
y
0 0.5242 0.1053 0.2009 8.0887 0.0269 0.1581 5.8783
1 0.6845 0.1430 0.2089 7.5574 0.0288 0.1604 5.5716
2 0.7031 0.1449 0.2061 7.7387 0.0281 0.1596 5.6765
3 0.7418 0.1728 0.2330 6.2443 0.0348 0.1674 4.8045
4 0.7520 0.1763 0.2344 6.1747 0.0352 0.1679 4.7634
5 0.7688 0.1928 0.2508 5.4756 0.0397 0.1728 4.3481
10 0.8499 0.2435 0.2865 4.2973 0.0506 0.1840 3.6340
FIGURE 17.11 Normalized compressive strength versus clay content for pultruded E glass/
MMT/PP.
~
33%
39%
48%
67%
87%
122%
0
0.5
1
1.5
2
2.5
N
o
r
m
a
l
i
z
e
d

c
o
m
p
r
e
s
s
i
v
e

s
t
r
e
n
g
t
h
0% 1% 2% 3% 4% 5% 10%
modulus. From these charts it is apparent that a significant improvement in com-
pressive strength (87%) and modulus (99%) is attainable for E glass/PP systems
with only 5% nanoclay content while using purely mechanical (unoptimized) means
of nanoclay dispersion, namely, a single-screw extruder and sonication at the die.
Examination of TEM micrographs of pultruded samples indicates the presence of
unexfoliated clay agglomerates, which when fully exfoliated would presumably yield
even better mechanical properties. As evidenced in Fig. 17.11, any negative effect
of the ultrasonic treatment on fiber alignment is small compared with the large
increase in matrix strength due to the presence of nanoclay particles. Further the
test data indicate that it is feasible to pultrude nanoclay-reinforced E glass/PP
laminates with up to 27% fiber volume fraction and up to 10% nanoclay content
without a dramatic increase in resin viscosity that could clog the consolidation die
or result in poor wetting of the fibers.
17.5.2 INELASTIC KINKING ANALYSIS FOR PURE
COMPRESSION LOADING
Microbuckling under axial compression of unidirectional composites is a failure
mechanism in which the fibers undergo concurrent kinking in a narrow band, as
depicted in Fig. 17.13. E is the angle of the kink band relative to the transverse
direction, I is the initial fiber misalignment angle, and I is the additional rotation
of the fibers in the kink band at critical stress. As mentioned in Section 17.1, Equation
17.1 is valid for elasticperfectly plastic matrix with kink band angle equal to zero,
which is an unrealistic assumption for the E glass/PP test specimens under consid-
eration. Thus Budiansky and Fleck
4
modified Equation 17.1 to include (1) matrix
FIGURE 17.12 Normalized compressive modulus versus clay content for pultruded E glass/
MMT/PP composite.
110%
99%
92%
72%
69%
62%
0
1
1.5
2
N
o
r
m
a
l
i
z
e
d

m
o
d
u
l
u
s

o
f

e
l
a
s
t
i
c
i
t
y
0% 1% 2% 3% 4% 5% 10%
2.5
0.5
strain-hardening assuming a Ramberg-Osgood relation and (2) kink band angle E !
0. The modified kinking equation takes the form
(17.2)
where R is a constant defining the eccentricity of the yield ellipse, i.e., R V
TY
/W
Y
,
and n is the strain hardening parameter. Based on the work presented in Reference
4, a step-by-step procedure for obtaining the critical shear strain (J ) within the kink
band is developed and presented in the following section. The objective of this
analysis is to be able to predict the change in the critical shearing strain in the kink
band (assuming steady state) with increasing nanoclay loading in the matrix.
For kink band angle E ! 0, from equilibrium of the kink band in steady state,
the applied longitudinal compressive stress (Fig. 17.13) on the composite can be
expressed in terms of transverse stress (V
T
) and shear stress (W) in the kink band as
(17.3)
With the assumption that far field shearing strain J
= 0, and that the additional

rotation of fibers I is positive, Equation 17.3 yields
4
J I (17.4)
The transverse normal strain in the kink band is given by
e
T
J tanE (17.5)
FIGURE 17.13 Kink band geometry and parameters.
+
Fibre
Kink band

c
n
Y
G
R
n
n
R
+
+ ( )
+
,
,
1
1 3 7
1
1
2 2
1
2 2
tan
/
/
tan
/ ]]
]
]
( ) n n 1 /
+
+
T
tan

Defining an effective shear strain,
(17.6)
Substituting Equation 17.5 in Equation 17.6 yields
(17.7)
Also, it can be shown that kink band shear stress
(17.8)
And kink band transverse stress
(17.9)
Hence the numerator of Equation 17.3 can be expressed as
(17.10)
Defining a constant parameter Equation 17.3 can then be
written as
or

e T
R e +
2 2 2

e
R + 1
2 2
tan

j
(
,
\
,
(
e
e

T
e
e
R
j
(
,
\
,
(
tan
2

+
j
(
,
\
,
(
+
j
(
,
\
,
(

T
e
e
e
e
e
R tan tan
2 2
1++ R
2 2
tan
+ 1
2 2
R tan ,
+

+
e e
e
+
j
(
,
\
,
(
G
e y
e
y y
/
/
1

+
2
G
e y
e
y y
/
/

giving
(17.11)
where G
*
D
2
G, and J
y
*
J
y
/D. As mentioned earlier, the elasticplastic strain
hardening behavior of the polypropylene matrix is modeled using a Ramberg-Osgood
relation given by Reference 4:
(17.12)
where n is the hardening parameter. Substituting Equation 17.12 in Equation 17.11
gives
(17.13)
In order to find the value of W
e
/W
y
for which Equation 17.13 reaches an extremum,
the derivative of Equation 17.13 with respect to W
e
/W
y
was obtained and set equal to
zero:
yielding the condition in the kink band
(17.14)
Substituting Equation 17.14 in Equation 17.12, an analytical expression for the
effective shearing strain at kinking (J
e
) is obtained:
(17.15)

+
G
e y
e
y y
/
e
y
e
y
e
y
n
+
j
(
,
\
,
(
3
7

+
j
(
,
\
,
(
+
G
e y
e
y
e
y
n
y
/
/
3
7
j
(
,
\
,
(
j
(
,
\
,
(

+
j
(
,
\
,
(
G
n
e
y
e
y
e
y
1
3
7
+
j
(
,
,
\
,
(
(
+
j
(
,
\
,
(
+
j
(
,
,
\
,
1
3
7
n
e
y
e
y
n
y
((
(
+
+
j
(
,
\
,
(
+
2
1
3
7
0
e
y
e
y
n
y

e
y
y
n
n
,
]
]
]
7
3 1
1
( / )
( )
/

e
y
y
n
y
n n
,
]
]
]
+

7
3 1 1
1
( / )
( )
/
( )
/

Finally combining Equation 17.5 and Equation 17.6 with Equation 17.15, the critical
steady-state shear strain in the kink band (J ) can be obtained:
(17.16)
Based on data reported by Budiansky et al.,
4
the values assumed for the constant
parameters are R 2, E 20q, and n 3. The kink band angle (E ) measured from
failed composite specimens was roughly around 25q. It should be noted that the
compressive strength of the composite is not very sensitive to the strain hardening
parameter.
4
The predicted values for the yield strain in shear (J
y
) and critical shear
strain in kink band (J ) as functions of nanoclay loading are tabulated in Table 17.3.
17.5.3 CALCULATION OF FIBER VOLUME FRACTION, COMPOSITE
SHEAR MODULUS, AND I/J*
y
RATIO
From stoichiometric analysis, the fiber volume fraction (V
f
) in the pultruded E glass/
PP composite was found to be 27.16%. It should be noted that Budiansky and Flecks
formula (Equation 17.2) was originally applied to composites with high fiber volume
fraction (~60%). The following stepwise procedure was used in calculating the compo-
site shear modulus and the I/J*
y
ratio:
1. Modulus of elasticity of the resin, E
m
, was calculated using the
stressstrain data collected from resin compression tests for different
percentages of clay loadings.
2. The Poissons ratio, X, of neat PP resin (0% clay loading) was 0.47.
11
It is assumed that Poissons ratio does not change with increasing clay
loading.
3. The shear modulus (G
m
) of the resin was calculated using the isotropic
formula
4. From Rosens formula, the shear modulus of the composite is related to
the matrix shear modulus through
where v
f
volume fraction of the fiber (~27%).
5. V
c
was obtained from composite compression test data reported in Table
17.3.
+

,
,
]
]
]
+
Y y
n
y
R
n
1
7
3 1
2 2
1
tan
( / )
( )
/
/

j
(
,
,
\
,
(
( ( ) n 1
G
E
m
m
+ ( ) 2 1
G
G
v
m
f
1
6. Knowing G
m
and V
f
, G was calculated, as reported in Table 17.3.
7. Knowing V
c
/G for a given nanoclay loading, the I/J*
y
ratio can be calcu-
lated using Equation 17.2, from which critical shear strain in kink band
(J ) can be calculated using Equation 17.16, as reported in Table 17.3.
8. For neat PP matrix, shear strain at yield is J
Y
0.0269.
11
An initial fiber
misalignment of I
-
0.2176 radians (~12.5q) was obtained using Equation
17.2 for the 0% clay loading case. The somewhat high fiber misalignment
angle is likely due to the fact that prepreg tapes were used instead of yarns
in the pultruder. It was assumed in all subsequent analysis that initial
misalignment I
-
remains unchanged with increased clay loading.
17.5.4 DISCUSSION OF RESULTS
To be able to better visualize the effect of nanoclay on compressive strength and
shear modulus of E glass/PP composite, experimental data for V
c
versus G from Table
17.3 are plotted in Fig. 17.14 for different nanoclay loadings. It is observed that both
compressive strength and composite shear modulus exhibited monotonic increases
with increasing nanoclay loading, as indicated by the symbols. In addition, consid-
ering that the line
-
IJ
y
0 corresponds to the Rosen solution for elastic buckling,
the fact most of the data points are located between
-
IJ
y
4 and
-
IJ
y
8 indicates
that the compressive failure for the nanoclay-reinforced pultruded E glass/PP is
definitely inelastic and that the formation of the kink band entails elasticplastic
deformations for this material system. Figure 17.15 shows values of matrix yield
FIGURE 17.14 Compressive strength versus composite shear modulus plot showing test data
and elastic-strain hardeningplasticity predictions for E 20q.
0.00
0.05
0.10
0.15
0.20
0.25
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
G
COMPOSITE
(GPa)
C
R
I
T
I
C
A
L
(
G
P
a
)
8
1 2 4
y
= 0
0% Clay
4% Clay
1% Clay
5% Clay
2% Clay
3% Clay
10% Clay
strain in shear (J
Y
) plotted as a function of clay loading calculated using Equation
17.2, following the stepwise procedure described in the previous section. Again a
significant increase (88%) in matrix yield strain in shear with nanoclay loading
was observed. Figure 17.16 depicts values of the critical shear strain in the kink
band (J ) plotted as a function of clay loading that was predicted using Equation
17.16. An approximately 17% increase in critical shear strain in the kink band
was predicted at 10% nanoclay loading, presumably due to the presence of nan-
oclay in the resin matrix. Finally Fig.17.17 depicts values of the ratio of critical
shear strain in the kink band (J ) to matrix yield strain in shear (J
Y
) plotted as a
function of clay loading. The fact that this ratio remains well above 1 for all
nanoclay loadings corroborates our earlier observation that the compressive failure
process for pultruded E glass/PP is inelastic, even though the degree of inelasticity
appears to decrease with increasing nanoclay loading.
FIGURE 17.15 Change in matrix shear strain at yield with nanoclay loading.
FIGURE 17.16 Change in shear strain in kink band with nanoclay loading.
88%
48%
31% 30%
5%
7%
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
0% 1% 2% 3% 4% 5% 10%
% of clay
N
o
r
m
a
l
i
z
e
d
y
0.90
0.95
1.00
1.05
1.10
1.15
1.20
0% 1% 2% 3% 4% 5% 10%
% of clay
N
o
r
m
a
l
i
z
e
d

The determination of compressive strength of pultruded composites is of consid-
erable importance, as these materials are generally weak in compression. This area
is one of the least understood fields in composites, especially nanocomposites,
because of the various parameters that affect the compressive behavior as well as
difficulty in determining the strength experimentally. Unidirectional E glass fiber-
reinforced pultruded composites were prepared in which the matrix was polypro-
pylene with layered silicate nanoclay reinforcement. Thermoplastic pultrusion was
used to prepare the composites. The compressive properties exhibited significant
increases with increasing clay loadings (up to 10 wt%). Continuum-based analytical
modeling was employed to predict the increase in critical shear strain in the kink
band assuming elastic strain hardeningplastic matrix behavior and nonzero kink
band angle. The model was able to establish that the kinking mechanism in E
glass/PP nanocomposite is definitely inelastic. However, the current model does
not explain the exact strengthening mechanism of the nanoclay reinforcement.
Multiscale simulations of the nanoclaypolymerglass fiber interactions would be
necessary to adequately address this issue. TEM studies indicate a favorable inter-
action between the silicate clays and the glass fibers, which are also silicon based,
indicative of enhanced matrixfiber adhesion. The clays are also presumed to
decrease the CTE mismatch, significantly reducing residual stresses and leading to
higher quality laminates.
Additional characterization experiments to determine resin and composite shear
properties are currently underway for additional verification of compressive failure
data. Experimental verification of predicted shear strains within the kink band will
be attempted using digital image correlation techniques. Another important aspect
that needs to be studied is the effect of processing variables, such as pull speed,
preformer temperature profile, and ultrasound on the orientation and dispersion of
nanoclay platelets in the pultruded specimens.
FIGURE 17.17 Variation in critical shear strain:yield strain ratio in kink band with nanoclay
loading.
0
1
2
3
4
5
6
7
0% 1% 2% 3% 4% 5% 10%
% of clay
y
PROBLEMS
1. Describe in detail the difference between intercalation and exfoliation of
nanoclay platelets. Why are these stages important in the processing of
high-quality nanoclay composites?
2. Explain why unidirectional fiber-reinforced composites are much weaker
in compression than in tension. What is the most frequently encountered
failure mode in compression of unidirectional composites?
3. What are some of the limitations of using the modified Argon formula
(Equation 17.1) to predict compressive failure in a unidirectional fiber
composite?
4. What were the changes made to the modified Argon formula (Equation
17.1) by Budiansky and Fleck (Equation 17.2)? Explain why.
5. Make a list of all of the processing variables used in the thermoplastic
pultrusion process. Discuss the effects of ultrasonic vibrations at the
consolidation die on the pultruded composite.
6. Discuss in detail the effect of addition of nanoclay particles on the fol-
lowing composite properties:
a. Compressive strength
b. Compressive modulus
c. Matrix yield strain in shear
d. Critical kinking strain
ACKNOWLEDGMENTS
This work was financially supported through a NASA-EPSCoR Research Initiation
Grant and through a seed grant from the Oklahoma NanoNet. Special thanks are
due to Dr. Shamsuzzoha at University of Alabama for his kind cooperation with
TEM characterization and Dr. Scott Taylor of Vybron Composites, Inc. for fabricat-
ing prepreg tapes. We also acknowledge Nanocor, Inc. for supplying us with the
surfactant-modified clay particles.
REFERENCES
1. Rosen, B.W., Mechanisms of composite strengthening, in Fiber Composite Materials,
American Society of Metals, Cleveland, OH, pp. 3775, 1964.
2. Argon, A.S., Fracture of composites, in Treatise on Material Science and Technology,
Vol. 1, Academic Press, New York, pp. 79114, 1972.
3. Budiansky, B., Micromechanics, Computers Struct., 16, 312, 1983.
4. Budiansky, B. and Fleck, N.A., Compressive failure of fiber composites, J. Mechan.
Phys. Sol., 41, 183211, 1993.
5. Creighton, C.J. and Clyne, T.W., The compressive strength of highly aligned carbon
fibre/epoxy composites produced by pultrusion. Composites Sci. Technol., 60,
525533, 2000.
6. Creighton, C.J., Sutcliffe, M.P.F., and Clyne, T.W., A multiple field image analysis
procedure for characterization of fibre alignment in composites, Composites Part A,
32, 221229, 2001.
7. Lu, H., Roy, S., Sampathkumar, P., and Ma, J., Characterization of the Fracture
Behavior of Epoxy Nanocomposites, Proceedings of 17th Annual Technical Confer-
ence, American Society for Composites, West Lafayette, IN, October 2123, 2002
(CD-ROM).
8. Taylor, S. and Thomas, W., High Speed Pultrusion of Thermoplastic Composites,
Proceedings of 22nd International SAMPE Technical Conference, November 68,
1990, Boston, MA, 1990.
9. Carsson, A. and Astrom, B.T., Experimental investigation of pultrusion of glass fiber
reinforced polypropylene composites, Composites Part A, 29A, 585593, 1998.
10. Liskey, A.K., Pultrusion on a fast track, Advan. Mater. Proc., 135, 2, 3135, 1989.
11. Maier, C. and Calafut, T., Polypropylene: The Definitive Users Guide and Databook,
Plastics Design Library, p. 268, 1998.
18
Nanomechanics
Young W. Kwon
CONTENTS
18.1 Introduction
18.2 Static Atomic Model
18.3 Coupling Atomic and FEA Models
18.4 Fatigue Analysis at Atomic Level
18.5 Heterogeneous Carbon Nanotubes
18.6 Summary
Problems
Acknowledgments
References
18.1 INTRODUCTION
As nanoscale devices and structures such as nanowires and nanotubes are being
designed and developed, it is important to analyze such devices and structures using
nanomechanics. There are multiple techniques for analyses of nanoscale structures.
Ab initio simulation methods and classical molecular dynamics are examples of
those techniques. The ab initio or rst-principles method is a numerical scheme to
solve rigorous quantum many-body Schrdinger equations and can deal with a very
small number of atoms. On the other hand, classical molecular dynamics are based
on Hamiltons classical equation of motion and can model a relatively large number
of atoms. Tight-binding molecular dynamics fall between the ab initio technique
and classical molecular dynamics. The tight-binding technique is based on simpli-
cation of the full quantum many-body problem so that its accuracy and the number
of atoms to be analyzed practically lie between those of the ab initio and classical
molecular dynamics techniques. Those simulation techniques are useful for design
and development of nanostructures.
Four different topics at the nanoscale are presented here: the static atomic model,
coupling between the discrete atomic model and the continuous nite element model,
fatigue analysis at the atomic level, and carbon nanotubes. The static atomic model
is benecial for predicting static properties of atomic structures such as Youngs
modulus. Furthermore coupling between the atomic model and a continuous medium
will be useful for multiscale analysis ranging from nanoscale to macroscale. The
fatigue analysis at the atomic level is to provide insight of fatigue damage processes
at the nanolevel, leading to eventual prediction of fatigue, life cycle, and design of
tougher materials. Finally heterogeneous carbon nanotubes are discussed, which
may be useful in future nanotechnology development.
18.2 STATIC ATOMIC MODEL
Classical molecular dynamics are based on Newtons second law of atomic motions.
The forces applied to atoms are derived from the energy potential among the atoms.
As a result, the theory of classical molecular dynamics solves second-order differ-
ential equations of motion of atoms using a numerical algorithm, i.e., using a time
integration scheme. There are many different numerical time integration techniques.
However, the two most popular techniques have been Varlets algorithm and Gears
predictor-corrector algorithm.
The classical molecular dynamics model computes positions, velocities, and
accelerations of atoms as a function of time. Then static and dynamic properties are
computed from the time history of the solutions. This is a time-consuming process.
Some properties such as elastic modulus can be determined more quickly through
static analysis without time integration of equations of motion. As a result, the
discrete static atomic model
1
is presented here for static analysis.
The discrete atomic model is based on static equilibrium of interatomic forces
among atoms. Interatomic forces are computed from the interatomic potential energy
expression. There are different mathematical forms of potential energies depending
on the bonding types of atoms. Simple potential expressions consider two-body
interaction, i.e., interaction between two atoms, without considering effects of other
atoms. More complex potentials include multibody interactions. Some of the poten-
tials are provided next.
One of the old and commonly used potential expressions is the LennardJones
potential:
2
(18.1)
where
(18.2)
Here (r
ij
) is the pair potential between two atoms i and j with distance r
ij
. In
addition, is the distance to the zero potential and is the energy at the minimum
of the potential energy. For Londons theory, which is one of van der Waals forces,
n 12 and m 6. Then the rst term models the short-range repulsion, while the
( ) r
r r
ij
ij
n
ij
m
j
(
,
\
,
(

j
(
,
\
,
(
,
,
,
]
]
]
]
j
(
,
\
,
(
n
n m
n
m
m n m /( )
second term indicates the attractive potential. The LennardJones potential has been
applied to amorphous solids or uids.
The Morse potential is as simple as the LennardJones potential, and it is
expressed in terms of exponential functions:
(18.3)
in which D and are constants with dimensions of energy and reciprocal distance,
respectively. In addition, r
o
is the equilibrium distance of the two atoms with (r
o
)
D, as shown in Equation 18.3.
The Gaussian-type potential is also a pair potential between two atoms such as
the following:
(18.4)
in which A, B, , and are material constants.
In the embedded atoms method,
35
which has been used for metallic atoms, the
energy of a collection of atoms is determined by the electron density, i.e., the
distribution of electrons in the system. As a result, the energy consists of two parts:
the energy to embed each atom into the electron density of the neighboring atoms
and the energy by a short-range pair interaction for the corecore repulsion. The
energy of a system can be represented by
(18.5)
where
h,i
is the total electron density seen by atom i due to the surrounding atoms
in the system, E
ee
is the embedding energy for placing an atom into that electron
density, and
ij
is the short-range pair interaction for the corecore repulsion between
atoms i and j separated by distance r
ij
.
In Equation 18.5, the total electron density seen by an atom is computed by
summation of atomic densities in surrounding atoms. That is,
(18.6)
in which
a
j
is the electron density contributed by atom j with distance r
ij
from atom i.
The short-range, repulsive pair potential used in References 35 is expressed as
ij
(r) = Z
i
(r)Z
j
(r)/r (18.7)
( )
( ) ( )
r D e e
ij
r r r r
ij o ij o

,
]
]
2
2

( ) r Ae Be
ij
r r
ij ij

2 2
+

E r
ee h i
i
ij ij
i j
( ) ( )
,
,

1
2

h i j
a
j i
ij
r
,
( )

The function Z(r) is determined by tting to the universal binding function.
The AbellTersoffBrenner potential
6
for carbons and hydrogens is expressed as
(18.8)
where
(18.9)
(18.10)
(18.11)
(18.12)
(18.13)
(18.14)
Here is the angle between the lines connecting atoms ij and ik. The constant
values in these equations are provided for carbon, hydrogen, and hydrocarbons in
Reference 6.
The force acting on atoms is computed from
(18.15)
in which is the unit position vector between the two atoms. The forces among N
atoms along with any external load in a given system must be in equilibrium.
Consider any two atoms located at positions i and j, called atom i and atom j,
as shown in Fig. 18.1. In that gure, solid circles indicate present positions of the
two atoms under equilibrium before applying external loads. As external loads are
applied to the system, the two atoms move to the positions denoted by open circles,
as seen in Fig. 18.1. Then displacement vectors from initial positions to nal positions
of the atoms are expressed as and , respectively. The position vectors of the two
atoms at the initial and nal positions, respectively, are denoted by and , as
( ) ( ) ( ) r V r B V r
ij R ij ij A ij

)
V r f r D S e
R ij ij ij ij
e
ij
S r
ij ij ij
( ) ( ) ( )
( )
(

1
2 RR
ij
e ( )
)
)
V r f r D S S e
A ij ij ij ij
e
ij ij
S
ij ij
( ) ( ) ( )
( )
/ (

1
2 rr R
ij
ij
e
( )
)
f
r R
r R
R R
ij
ij ij
ij ij
ij ij
1
1
1
1
2
( )
( )
( ) (
cos
(
11
1 2
2
0
)
( ) ( )
.
j
(
,
\
,
(
,
,
,
]
]
]
]
< <
R r R
r R
ij ij ij
ij ijj
( ) 2
B B B F N N N
ij ij ji ij i
t
j
t
ij
conj
+ +
( )
( )/ , ,
( ) ( )
2
B G f r e
ij i ijk ik ik
r R r
ijk ij
ij
e
i
+

1 ( ) ( )
( ) (
( )

jj
ij
e
R
ij i
H
i
C
k i j
H N N
+
( )
,
,
,
]
]
( )
)
( ) ( )
( , )
,
]]
]
i
G a
c
d
c
d
o
o
o
o
o
( )
( cos )
+
+ +

,
,
,
]
]
]
]
1
1
2
2
2
2 2
F r
r
n
ij
ij
r
( )

r
n
r
u
i
u
j
r
ij
R
ij
shown in Fig. 18.1. The displacement vectors and position vectors are related as
(18.16)
where is the relative displacement vector of the two atoms.
The force between the two atoms i and j at their new equilibrium positions is
expressed as Equation 18.17,
(18.17)
in which R
ij
is the distance between the two displaced atoms. is the directional
unit vector along the vector and it is expressed as
(18.18)
Equation 18.18 is substituted into Equation 18.17 and the resultant expression is
(18.19)
Applying Equation 18.19 to atoms i and j yields a matrix expression
(18.20)
FIGURE 18.1 Positions of two atoms before and after displacements.
z
u
i r
ij
R
ij u
j
Atom j
y
x
Atom i
Atom j
Atom i
R r u u r u
ij ij j i ij ij
+ +
u u u
ij j i

F R F R n
ij ij ij ij R
( ) ( )
n
R
R
ij
n
r u
R
R
ij ij
ij
+
F R F R
u
R
F R
r
R
ij ij ij ij
ij
ij
ij ij
ij
ij
( ) ( ) ( ) +
k k
k k
k k
k k
k k
k k
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
,
,
,
,
,
,
,
,
]
]
]
]
]
]
]
]
]
]
j
(
,
,
,
u
u
u
u
u
u
ix
iy
iz
jx
jy
jz
,,
,
,
,
\
,
(
(
(
(
(
(
(
j
(
,
,
,
,
,
,
,
\ F
F
F
F
F
F
ix
iy
iz
jx
jy
jz ,,
(
(
(
(
(
(
(
where
(18.21)
(18.22)
(18.23)
(18.24)
Here (x
i
, y
i
, z
i
) and (x
j
, y
j
, z
j
) are the coordinates of the atoms, as located in
Fig. 18.1.
The matrix expression in Equation 18.20 is computed for all atoms that interact
with one another and is assembled into the system matrix consisting of all atoms
displacements. This resultant system matrix equation is nonlinear and it is solved
after applying constraints of atoms. Because and are not known a pri-
ori, , , and are assumed initially or obtained from the previous iteration
solutions with a proper relaxation technique.
Figure 18.2 shows initial equilibriums of atoms in square arrays with a disloca-
tion oriented at 27, 45, and 63, respectively, to the horizontal axis. The atoms at
the bottom row are constrained from any movement, while the atoms at the top row
are subjected to uniform tensile loading. As the tensile load is applied, the displace-
ments of atoms into the new equilibrium in each case are depicted in Fig. 18.3. This
gure illustrates how a slanted dislocation affects the movement of atoms under a
uniform load.
Figure 18.4 shows two different types of carbon nanotubes: armchair and zigzag
shape. The armchair model is a single-wall nanotube of the chiral vector of (6,6),
while the zigzag single-wall nanotube has the chiral vector of (12,0). Using the
AbellTersoffBrenner potential function,
6
the equilibrium state of the initial arm-
chair model is computed. After the atomic equilibrium state, the axial tensile force
was applied to some atoms in the longitudinal direction of the tubes. As a result of
the force, the nanotubes had elongations. Then the effective stiffness of the nanotubes
was calculated as
(18.25)
where is the applied stress and l/l where l is the elongated length of tube
and l is initial length of the tube. The computed elastic modulus of zigzag nanotube
was 1.4 TPa, which compared well with other available data.
k
F R
R
ij ij
ij
( )
F F
F R
R
x x
ix jx
ij ij
ij
i j

( )
( )
F F
F R
R
y y
iy jy
ij ij
ij
i j

( )
( )
F F
F R
R
z z
iz jz
ij ij
ij
i j

( )
( )
R
ij
F R
ij ij
( )
u
i
u
j
R
ij
E


18.3 COUPLING ATOMIC AND FEA MODELS
The atomic model is a discrete system and it is useful to couple the atomic model
to a continuum model. A practical size of the discrete system for computational
modeling is very small because atoms are separated by distances one-tenth on the
order of a nanometer. On the other hand, a continuum can be modeled in any size.
Therefore coupling of the discrete atomic and continuous models provides a unique
benet. In other words, a relatively large size of domain can be analyzed in con-
junction of accurate representation of atomic behavior, which is not obtainable in a
continuous model.
FIGURE 18.2 Square arrays of atoms with dislocations: (a) 27, (b) 45, (c) 67 orientation
from the horizontal axis.
25
20
15
10
5
0
5
10
15
20
25
25 20 15 10 5 0 5 10 15 20 25
(a)
25
20
15
5
10
0
5
10
15
20
25
25 20 15 10 5 0 5 10 15 20 25
(b)
25 20 15 10 5 0 5 10 15 20 25
25
20
15
10
5
0
5
10
15
20
25
(c)
There are different approximation techniques to analyze continuous systems. One
of the most common techniques is the nite element method. As a result, the coupling
technique is described here between the nite element model and the atomic model.
If another method such as the boundary element, nite difference, or meshless tech-
nique is used, the overall procedure remains the same except for some modications
in details, as necessary.
For simplicity, consider a two-dimensional coupling between the two models
even though the same algorithm can be applied to three-dimensional bodies.
Figure 18.5 shows discrete atoms surrounded by a continuous medium discretized
FIGURE 18.3 Movements of atoms with dislocations subjected to vertical tensile loads: (a) 27,
(b) 45, (c) 67 orientation from the horizontal axis.
25 20 15 10 5 0 5 10 15 20 25
30
20
10
0
10
x-axis
y
-
a
x
i
s
20
30
(a)
30
20
10
0
10
20
30
25 20 15 10 5 0 5 10 15 20 25
x-axis
y
-
a
x
i
s
(b)
25 20 15 10 5 0 5 10 15 20 25
30
20
10
0
10
20
30
x-axis
y
-
a
x
i
s
(c)
for a nite element mesh. In the gure, the inner domain is the atomic domain,
while the outer domain is the nite element domain of the continuous medium. The
intermediate domain bounded by bold lines where both atoms and nite element-
meshes overlap each other is called the interface domain for coupling. The atoms
FIGURE 18.4 (a) Armchair and (b) zigzag carbon nanotubes.
FIGURE 18.5 Coupling between discrete atomic model and continuous FEA model.
35
30
25
20
15
10
5
0
60
50
40
30
20
10
0
6
6
6
4
4
4 4
2
2
2
2
0
0
6
4
2
2
4
6
6
4
2
0
0
2
(a) (b)
Discrete atomic
domain
Interface domain
for coupling
Continuum domain with
finite element mesh
in the interface domain are called embedded atoms. The solution procedure with
staggered solutions between the atomic domain and the nite element domain is as
follows:
1. Solve the nite element matrix equation of the continuous
domain. From now on, subscripts f and a are used to denote nite element
and atomic domains, respectively.
2. Compute the embedded atoms displacements in the interface domain
from the nite element nodal displacements using the nite element shape
functions such as , where superscript e indicates the embed-
ded atoms and nite elements in the interface domain, and [N] is the
shape function matrix of nite elements in the interface domain.
3. Compute the new positions of the embedded atoms by adding the
displacements computed in step 2 to the previous positions.
4. Using the atomic model, solve for the new positions of the atoms in the
atomic domain with xed embedded atoms positions.
5. Compute the interactive forces on the embedded atoms exerted by
all atoms.
6. Compute the equivalent nodal forces of the nite elements at the interface
using where V
e
is the nite element volume.
7. With the nodal forces computed in step 6, the new nite element solution
is obtained. Then continue the process from Step 2.
Figure 18.6 shows a coupling between the atomic domain and a nite element domain
with a crack. The close-up view of the atomic domain of Fig. 18.6 is shown in
Fig. 18.7. In this case, boundary conditions such as external loading and constraints
can be applied to the nite element model. The problem domain was subjected to a
tensile load perpendicular to the crack orientation at the top boundary, while the bottom
FIGURE 18.6 Coupling between atomic model and nite element model with a crack.
[ ]{ } { } K u F
f f f
{ } [ ]{ } u N u
a
e
f
e
{ } x
a
e
{ } F
a
e
{ } [ ] { } F N F dV
f
e T
V
a
e
e

150
150 200
100
100
50
50
0
0
50
100
100 50
150
boundary was xed. The deformed nite element domain is plotted in Fig. 18.8. The
associated atoms movements are plotted in Fig. 18.9 with the atoms nal positions
shown in Fig. 18.10.
18.4 FATIGUE ANALYSIS AT ATOMIC LEVEL
Most structural materials, especially metals, fail due to repeated loads. This is called
fatigue failure. Fatigue is a damage accumulation process in materials under repeat-
ing loads. Material defects at the atomic level, such as impurities, inclusions, vacan-
cies, and dislocations, play an important role in fatigue process. When a material is
FIGURE 18.7 Close-up view of the atomic domain of Figure 18.6.
FIGURE 18.8 Deformed nite element domain with vertical tensile loading of Figure 18.6.
60
60 80 100 120
40
40
20
20
0
0
20
20
40
60
x-axis
y
-
a
x
i
s
150
150 200
100
100
50
50
0
0
50
100
100 50
150
stressed even in the elastic range in macroscale, there are irreversible changes within
the material that are so small that they are not measurable in the stress-strain curve.
However, as the number of cycles in the loads increases signicantly, the accumu-
lated damage leads to ultimate failure of the materials.
7
In the past, prediction of fatigue failure mainly relied on experimental observation
and data, which were time-consuming processes. Fatigue failure is inuenced by many
different parameters. Considering all the parameters affecting fatigue failure in exper-
iments is impossible and practically prohibitive. Furthermore, because nanotechnology
FIGURE 18.9 Atomic displacements associated with Figure 18.7.
FIGURE 18.10 Atomic positions associated with Figure 18.7.
60
40
40 60 80 100 120
20
20
0
0
20
20
40
60
80
60
60 80 100 120
40
40
20
20
0
0
20
20
40
60
has been developed recently, many devices have been devised at nano- or at most
microscale. Even those scale specimens were shown to fail as a result of fatigue.
8
Experimental testing for fatigue behavior at nano- or microscale is quite a challenge.
Analytical modeling and simulation of fatigue processes are very benecial to
alleviate the heavy burden of experimental work and to enhance the understanding
of fatigue processes. The present task is to utilize the classical molecular dynamics
to understand the fatigue damage mechanisms in metallic materials.
A pure metal such as copper was studied using the embedded atom method.
3,4
The atomic structure with dislocation is shown in Fig. 18.11. The atomic arrange-
ment was determined after applying a number of cyclic loads, as described later.
For visual clarity, dislocation was placed through the z-axis and the xy-plane view
is depicted in the gure. The atoms on the left face were constrained from any
movement and a cyclic axial loading was applied uniformly to atoms on the right
face under an isothermal condition. The applied cyclic loading (force versus. time)
is shown in Fig. 18.12 at a constant room temperature. The applied peak stress was
100 or 200 MPa and the period of load cycle was 200 times the time step size.
FIGURE 18.11 Plane view of copper atomic arrangement with dislocation.
FIGURE 18.12 Cyclic loading.
35
30
25
20
15
10
10 15 20 25 30 35
X
Y
Force
Time
Because of the cyclic load, there is a change of potential energy in the copper
atoms. Figure 18.13 and Fig. 18.14 plot the average potential energy per copper
atom as a function of time as the cyclic loading continues. There are uctuations of
the average potential energy, as seen in the gures. However, linear curve tting of
the uctuations indicates the increase of the average potential energy along with
cyclic loading. The increase of the average potential energy is larger in Fig. 18.14
than in Fig. 18.13 because a larger stress is applied to the system in Fig. 18.14. With
the linear curve tting, the slope of the curve may predict when the separation of
atoms occurs, i.e., the fatigue failure of the material. In other words, the life cycle
(number of cycles to failure) can be estimated from the rate of potential energy
increase per cycle and the potential energy change necessary for atomic separation.
In order to explain why there is an increase in potential energy, a simple atomic
model was presented. Let us consider an interaction between two atoms whose
interaction is described in terms of a nonlinear potential function, for example, Morse
potential for mathematical simplicity. One of the two atoms is xed, while the other
vibrates. Assuming one-dimensional motion, the dynamic equation of the vibrating
atom can be written as
(18.26)
where m is the atomic mass, x is the displacement of the vibrating atom from its
equilibrium position, and f(x) is the interaction force between the atoms, which is
FIGURE 18.13 Change of potential energy per atom resulting from a cyclic load between
0 and 100 MPa versus the number of time steps.
mx f x ( ) + 0
5
4
3
2
1
0
1
2
3
4
5
0 500 1000 1500 2000 2500 3000 3500 4000
No. of time steps
C
h
a
n
g
e

o
f

p
o
t
e
n
t
i
a
l

e
n
e
r
g
y
10
4
derived from the interaction potential. Rewriting the equation in terms of state
variables, displacement x
1
x and velocity yields
(18.27)
The plot of phase trajectory of Equation 18.6 indicates the stability of the system
and it depends on the initial conditions of the displacement and velocity.
Velocities of atoms under equilibrium obey the MaxwellBoltzmann law.
9
Considering a one-dimensional problem, the fraction of atoms having velocities
between v and v + dv can be expressed as
(18.28)
where k is Boltzmanns constant and T is the temperature. Therefore, some of the atoms
have much higher velocities than the average velocity of the system. Those atoms with
high velocities show unstable phase trajectory, as shown in Fig. 18.15, while atoms
with low velocities have stable phase trajectory, as shown in Fig. 18.16. The unstable
phase trajectory means an increase of position or increase of potential energy. In other
FIGURE 18.14 Change of potential energy per atom resulting from a cyclic load between
0 and 200 MPa versus the number of time steps.
10
5
0
5
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
No. of time steps
C
h
a
n
g
e

o
f

p
o
t
e
n
t
i
a
l

e
n
e
r
g
y
10
4
x x
2

( )/
x x
x f x m
1 2
2 1

g v dv
m
kT
mv
kT
dv ( ) exp
j
(
,
\
,
(
2 2
2

words, when a system of atoms is subjected to an external loading, some atoms at
unstable equilibrium will increase potential energy, thus contributing to the increase
of the total potential energy in the system.
Furthermore because not all atoms vibrate at the same frequency and amplitude,
the vibrating frequencies of some atoms can be near or at the excitation frequency
or frequencies of the applied repeated loading. Then the displacements of the atoms
increase, resulting in the increase of potential energy.
FIGURE 18.15 Unstable phase trajectory for an atom with an initial high velocity.
FIGURE 18.16 Stable phase trajectory for an atom with an initial normal velocity.
1.5
1
0.5
0
0 5 10 15
0.5
1
1.5
5
Displacement
V
e
l
o
c
i
t
y
1.5
1.5
1
1
0.5
0.5
0
0
0.5
0.5
1
1
1.5
Displacement
V
e
l
o
c
i
t
y
18.5 HETEROGENEOUS CARBON NANOTUBES
Conventional carbon nanotubes (CNTs) have uniform atomic structures along
their lengths as shown in Fig. 18.17. In order to have a nonhomogeneous arrange-
ment of atoms along the tube wall, a bamboo structural nanotube (BSNT) was
devised by combining two single-wall nanotubes (SWNTs), as seen in Fig. 18.18.
FIGURE 18.17 Initial (5,5) armchair SWNT model from geometrical consideration.
14
12
10
8
6
4
2
0
Z
-
a
x
i
s

(
)
Y-axis ()
2 0 2
(a) Side view of the initial SWNT model
3
2
2 3
1
1
0
0
1
1
2
2
3
3
X-axis ()
Y
-
a
x
i
s

(
)
(b) Top view of the initial SWNT model
Heterogeneous carbon nanotubes with specic purposes may be controlled and
shaped in the near future.
For the BSNT model, let us assume that there are two SWNTs as CNT A and
CNT B with the same diameter: CNT A has capped ends at both top and bottom,
while CNT B has the capped top and the open bottom. When the bottom of CNT B
is physically contacted with the capped top of CNT A along the tubes z-axis with
a revolved horizontal angle of 0.5 radian, CNT A and B can be jointed as a hetero-
junction region to lower the total system energy. When two nanotubes were coupled
with a horizontal joint angle of 0 radian, the heterojunctions could not be formed
during equilibrium simulation due to the chemical and geometrical bonding limits.
Figure 18.18 shows the side and top views of a BSNT model. The two (5,5)
armchair SWNTs, upper and lower SWNTs, were prepared for modeling BSNT with
the average CC distance 1.42 and diameter 6.78 . The resultant tube length of
the initial BSNT model was 15.14 measured from the origin coordinate (0,0,0)
at the bottom of the lower tube to the tip of the upper tube end along the z-axis.
The number of carbon atoms for the BSNT model is 210.
Since additional pentagon defect rings are introduced in the junction region,
the mechanical properties of the BSNT such as Youngs modulus will be different
from the previous SWNT model. This section describes the method to calculate
Youngs modulus of SWNT and BSNT by introducing the freestanding room
temperature vibration method presented in Reference 10. According to the kinetic
molecular theory, carbon atoms are vibrating about their equilibrium positions with
a mean vibrational kinetic energy that increases with the temperature as 3/2 kT,
where k is the Boltzmann constant and T is temperature. From this fundamental
principle, Krishnan et al.
10
estimated Youngs modulus of single-wall carbon nano-
tubes as Y 1.25 0.35/ + 0.45 TPa by observing a SWNTs free-standing
room temperature vibrations in a transmission electron microscope (TEM).
Their formula to calculate Youngs modulus was derived from the relationship
between the motion of a vibrating clamped cylindrical cantilever rod governed by
the classical fourth-order wave equation and the quantum mechanics statistical
probability theory given by the Boltzmann factor. The resultant relationship between
Youngs modulus Y, length L, inner and outer tube radii b and a, which form a
cylindrical CNT wall, and the root mean square (rms) of displacement , which is
represented by the vibration amplitude at the tip of a carbon nanotube at a temper-
ature T, can be expressed by
(18.29)
where W is the SWNTs width (diameter), G( a b) is the graphite interlayer
spacing of 0.34 nm, and k is the Boltzmann constant.
For the present study, Equation 18.29 is used to predict Youngs modulus:
(18.30)
2
3
2 2
0 8486
+
.
( )
L kT
YWG W G
Y
L kT
WG W G
+
0 8486
3
2 2 2
.
( )
FIGURE 18.18 Side and top views of BSNT model with 210 atoms.
15
10
5
0
Z
-
a
x
i
s

(
)
Y-axis ()
2 0 2
(a) Side view of BSNT model
3
2
2 3
1
1
0
0
1
1
2
2
3
3
X-axis ()
Y
-
a
x
i
s

(
)
(b) Top view of BSNT model
From this equation, it should be noticed that Youngs modulus depends strongly on
the rms displacement at the tip of the SWNT for a given length L, width W of the
CNT, and the given ambient temperature T.
How far will a carbon atom at the tip of SWNT or BSWNT vibrate in a given
parameter such as time interval, diameter, and temperature? In the three-dimensional
Cartesian coordinate {x, y, z}, a position vector r
i
that locates carbon atom i in the
CNT is dened by three linearly independent vectors as
(18.31)
If the atoms are allowed to move from their equilibrium positions by an amount of
displacement vector at a time t
0
, then the actual position at the time
of any carbon atom under the inuence of a vibration (thermal uctuation) is given by
(18.32)
After a time interval t, the changed position of atom can be written as
(18.33)
From Equation 18.32 and Equation 18.33, the mean square displacement (msd)
of a carbon atom i at the tip of SWNT at given time t can be determined by
(18.34)
where the notation denotes averaging over all the atoms.
When this quantity is averaged over all N atoms consisting of the end tip of the
SWNT or BSNT, the rms displacement of the carbon nanotube-tip can be written as
(18.35)
In a molecular dynamics simulation to nd Youngs moduli for SWNT and BSNT
models under the equilibrium state, Equation 18.31 through Equation 18.35, played
an important role to calculate the numerical rms displacement values from the
simulated trajectory data of the carbon atoms, which were then applied in Equation
18.30 to compute Youngs moduli.
The two models generated previously were tested under the molecular dynamics
simulation at room temperature, 300 K, and then Youngs moduli of SWNT and
BSNT were evaluated based on Equation 18.30 and Equation 18.35. After that the
numerical values were compared with the theoretical and experimental results pre-
sented from other studies.
Figure 18.19 histograms demonstrate the spread in the evaluated Youngs moduli
for SWNT and BSNT by running 100 simulations for each model. The measured
r Xx Y y Zz
i
+ +
u u u u
i ix iy iz
( , , )
R t r t u t
i o i o i
( ) ( ) ( ) +
0
R t r t u t
i i o i
( ) ( ) ( ) +
msd t R t R t
i i i
( ) ( ( ) ( ))
0
2

j
(
,
,
,
,
,
\
,
(
(
(
(
(
msd t
N
i
i
N
( )
/
1
1 2
objects were the carbon atoms of the pentagon ring poles at the tips in both models.
The simulation code was designed to calculate the rms displacements of the tip
carbon atoms using Equation 18.31 through Equation 18.35 when the instantaneous
kinetic energy per atom of both models reached 0.039 eV (3/2 kT) at given room
temperature and at 300 K. Then Youngs modulus was calculated using Equation
18.30. In both cases, the use of 100 simulations was considered sufcient to reduce
statistical errors due to the random initial velocities for both models. The simulation
results are tabulated in Table 18.1.
FIGURE 18.19 Histograms of SWNT and BSNT Youngs modulus.
30
25
20
15
10
5
0
0.5
1 1.5 2 2.5 3 3.5
Y (Pa) 10
12
C
o
u
n
t
s
(a) Histogram of SWNT Youngs modulus
25
20
15
10
5
0
4 5 6 7 8 9 10
Y (Pa) 10
11
C
o
u
n
t
s
(b) Histogram of BSNT Youngs modulus
The mean value of Youngs modulus for the SWNT model was 1.42 TPa with
the standard deviation of 0.34 TPa. Based on this data, the expected mean Youngs
modulus for the real population of SWNT can be predicted as 1.42 0.23 TPa with
50% condence intervals. These values are in good agreement with the theoretical
and experimental Youngs modulus values reported in Reference 10.
In particular, considering about 48% of Youngs moduli for the SWNT model, which
were distributed within the range 1.0 1.5 TPa, and comparing with the previously
mentioned mean Youngs modulus of SWNT, 1.25 TPa as evaluated by the same
thermal vibration method presented by Krishnan et al.,
10
the consistency of both results
is comparable. In their method,
10
the tubes length and tip vibration amplitudes were
estimated directly from the digital micrographs (TEM images). From this study, the
molecular dynamics simulation was proved to be reasonable to evaluate Youngs modulus
of the SWNT model, and it can be also used to simulate the BSNT model.
The molecular dynamics simulation results of the SWNT and BSNT models
show two important features. One is that though present SWNT results and prior
studies are consistent with each other, the Youngs modulus value (mean 1.42 TPa)
of the present SWNT model calculated from the molecular dynamics simulation
shows a somewhat higher value than the Youngs modulus value (mean 1.25 TPa)
used with the same thermal vibration method reported by Krishnan et al.
10
The
second is that the calculated Youngs modulus value of the BSNT model shows an
obvious difference with respect to the SWNT model. The mean Youngs modulus
of the BSNT model was calculated as 0.604 TPa with a standard deviation 0.1 TPa.
Consequently the expected mean Youngs modulus for a BSNT population would
be predicted to be 0.604 0.07 TPa with a 50% condence level. The results show
a signicantly lower Youngs modulus of BSNT than that of SWNT. Unfortunately
there are thus far no available data for the heterogeneous CNTs Youngs modulus
to be compared with the value of the BSNT model. Nevertheless it might be expected
that this study will play a starting role to speed up the progress of heterogeneous-
type nantotube synthesis in further studies.
Why is the Youngs modulus of the SWNT model higher than that of the BSNT
model by roughly twice, as shown in Fig. 18.19? Earlier studies reported that the
Youngs modulus of CNTs depends on the presence of structural imperfections such
as the nesting of tubular cylinders, which can create a joint or knuckle, thereby
weakening the tube, and structural defects of the pentagon (5) and heptagon (7) rings.
TABLE 18.1
Statistical Data Measured from Molecular Dynamics
Simulations of SWNT/BSNT Model
Parameter SWNT BSNT
Number of simulations 100 100
Mean Youngs modulus (TPa) 1.4243 0.6041
Standard deviation (TPa) 0.342 0.1002
Maximum Y value (TPa) 3.4202 0.9241
Minimum Y value (TPa) 0.7402 0.4046
In the BSNT model shown in Fig. 18.17(b), the defects along the circumference can
prove the arguments presented by Srivastava et al. ,
11
who stated that heterogeneous
CNTs induced 57 defect rings, which absorbed partial energy from the total energy
of the system because it required defect formation energy.
Thus the defects at the junction regions of the BSNT model may absorb energy.
Then the energy needed for the atomic vibration in the BSNT tips would increase,
and Youngs modulus of the BSNT model would be lower than that of the SWNT
model based on the relationship of Equation 18.30. This analytical interpretation is
also consistent with the results of Fig. 18.19(b), as mentioned previously.
The deformation behaviors of SWNTs and BSNTs under an external tension
load were also simulated with the molecular dynamics code to compare Youngs
moduli of both models. When a system is subjected to an external load, the system
is forced out of its initial equilibrium into a new equilibrium with the given load.
Series of the molecular dynamics simulation plots at the nanoscale are presented to
investigate the mechanical behaviors and reactions of both models.
Figure 18.20 and Fig. 18.21 are molecular dynamics simulation plots of the
tensile loading test that were performed on both models at various forces of 30
60 (nN/atom), respectively. Figure 18.20(a) and Fig. 18.21(a) show the initial set-
up of the SWNT and BSNT after the equilibrium relaxation. The bottom atoms of
both models are constrained as if they are anchored at a substrate material. Then,
z-directional tensile forces were applied to the carbon atoms in the cap regions of
both SWNT and BSNT such as a conventional tensile specimen.
Figure 18.20(b) and Fig. 18.21(b) show the congurations of both models when the
applied tensile force is 30 (nN/atom), which means that both models are within the small
strain region. Although separation gaps have appeared in both models, their average
separation distances are less than the maximum potential inuence distance of 2
between carbon atoms. Beyond the scope of this tensile force, both carbon nanotubes
show characteristic buckling and kicking features at the side tubular wall. These features
are also shown in Fig. 18.20(c) and Fig. 18.21(c) where the applied force is 50 (nN/atom).
Figure 18.20(d) and Fig. 18.21(d) demonstrate that both SWNT and BSNT models
separate into two halves (top and bottom) beyond the maximum tensile force (60 nN/
atom), leading to a fracture, which means that the average separation gaps are larger
than 2 . This separation and fracture are not rapid, but occur gradually due to local
interactions or rearrangement between the carbon atoms. However, slip taking place
at about 45 to the loading direction, which is a general phenomenon in the case of
ductile metals, was not observed explicitly. Rather, under the strong tensile force,
they exhibited a brittle feature similar to a diamond, which is known as the hardest
material consisting of carbon atoms with an sp
3
bonding structure.
Figures 18.21(e) and (f ) show the changes in the cross-sectional area and
distribution of atoms seen from the top view of the BSNT model from the case of
initial conguration (e), to the case of large applied tensile force of 75 nN/atom (f ).
A precise value for the changed cross-sectional area is difcult to obtain, because
the number of distortions of circumferential carbon atoms in the side wall of the
BSNT model is not obvious. This unclear cross-sectional area that appears during
the molecular dynamics simulation will prevent computing the applied stress needed
for calculating Youngs modulus from molecular dynamics results. Furthermore
carbon atoms in the dome-shaped cap region of the BSNT model show the rotational
motions on the axis as the applied force increases. Both observations demonstrate
that carbon atoms under the potential and an external force look for the appropriate
position to minimize the system energy. Thus Fig. 18.21(f ) directly proves that the
binding between atoms can be formed or broken due to the induced defects on the
BSNT model during the simulation.
Next a relative Youngs modulus ratio concept for the SWNT and BSNT will
be used with the force-strain diagram extracted from the molecular dynamics sim-
ulation, which will then be compared with the Youngs modulus values calculated
from the previous simulation.
FIGURE 18.20 Congurations of the SWNT model under various magnitudes of tensile
loads: (a) no loading, (b) 30 nN/atom, (c) 50 nN/atom, (d) separation of atoms.
14
12
10
8
6
4
4
4
4
4
2
2
2
2
2
0
0
0
(a)
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
4 2 0 2 4
(b)
16
14
12
10
8
6
4
4 2 0 2 4
2
0
(c)
18
16
14
12
10
8
6
4
2
0
4
4
4
4
2
2
2
2
0
0
(d)
FIGURE 18.21 Congurations of BSNT model under various magnitudes of tensile loads:
(a) no loading, (b) 30 nN/atom, (c) 50 nN/atom, (d) separation of atoms, (e) top view of
initial atomic conguration, (f ) top view of atoms under a large tensile load.
15
10
5
0
4
4
4
4
2
2
2
2
0
0
(a) 16
14
12
10
8
6
4
4 4
2
2 0 2 4 2 0 2 4
0
(b)
16
14
12
10
8
6
4
4 4
2
2 0 24 2 0 2 4
0
(c)
18
16
14
12
10
8
6
4
2
0
4
-4
2
2
0
0
2
2
4
4
(d)
4
4
3
3
2
2
1
1
0
0
1
1
2
2
3
3
4
4
X-axis
Y
-
a
x
i
s
(e)
4
4
3
3
2
2
1
1
0
0
1
1
2
2
3
3
4
4
X-axis
Y
-
a
x
i
s
(f)
By conventional continuum mechanics, the ultimate tensile strength and Youngs
modulus for a bulk material can be determined from the force-displacement data
under the external tensile loading test. The ultimate tensile strength is measured as
the maximum stress prior to fracture. The modulus is measured as the slope within
the small strain limit range.
12
The tensile force-strain diagram for the SWNT and BSNT obtained from the
molecular dynamics simulation is plotted in Fig. 18.22. The applied tensile forces
per atom of the SWNT and BSNT were recorded on the y-axis, and the average
strains of carbon atoms in the location right below the forced atoms were recorded
on the x-axis. Overall the observed force-strain diagrams show a signicant nonlinear
relationship. In particular, it is worthwhile observing that the strain of BSNT is more
linearly increased than that of SWNT as the applied force increases. The plot shows
that the Youngs modulus of BSNT is less than the modulus of SWNT.
To calculate the relative Youngs modulus between two materials, SWNT and
BSNT, the linear regression analysis was applied as shown Fig. 18.23. Based on the
linear regression analysis, the ratio of force to strain of a specic material can be
represented as the slope of the regression line. In Fig. 18.23, while the data set of
FIGURE 18.22 Tensile force-strain per atom for the SWNT and BSNT models.
FIGURE 18.23 Linear regression tting for SWNT and BSNT models within elastic limits.
0
10
20
30
40
50
60
70
0 0.01 0.015
F
o
r
c
e

p
e
r

a
t
o
m

(
n
N
)
Strain
0.005 0.02
BSNT
SWNT
0
5
10
15
20
25
30
35
F
o
r
c
e

p
e
r

a
t
o
m

(
n
N
)
y(SWNT) = 7416.3x + 5.0718
R
2
= 0.9196
y(BSNT) = 3691.3x + 9.6543
R
2
= 0.9893
BSNT
SWNT
Linear (SWNT)
Linear (BSNT)
0 0.001 0.002 0.003 0.004 0.005 0.006
Strain
SWNT shows a more nonlinear characteristic than those of BSNT within the same
applied force range, the linear regression analysis can be used for both data sets due
to a good tting statistical coefcient R
2
, which quanties a goodness of t and is
a fraction between 0.0 and 1.0, with no unit. Higher R
2
values indicate that the tting
line or curve comes closer to the data. In this case, the computed R
2
using the
statistical software package was 0.91 for SWNT and 0.99 for BSNT. These numerical
values indicate that the linear assumption used to calculate the slope of a regression
line for SWNT and BSNT is appropriate.
The resultant slope of the regression line is 7416.3 for SWNT and 3691.3 for
BSNT, as seen in Fig. 18.23. Consequently the relative Youngs modulus ratio of
the BSNT with respect to the SWNT was calculated as 0.498. This means that the
average Youngs modulus of the BSNT model is 49.8% that of SWNT.
From the results of the previous equilibrium simulations, the evaluated mean
Youngs moduli of SWNT and BSNT were 1.424 TPa and 0.604 TPa, respectively.
Based on these two values, the ratio of Youngs modulus is 0.424 (42.4%). This is
consistent with the aforementioned value, 0.498 (49.8%), within an error of 15%.
Consequently Youngs modulus of the BSNT model was observed to be lower than
that of the SWNT model, indicating that heterogeneous CNTs have lower Youngs
moduli than pure CNTs. However, the heterogeneous CNT still has a larger modulus
than conventional metals. The heterogeneous CNTs may be great potential candi-
dates promising high-strength and ultra-lightweight composite materials and nano-
scale devices in the near future.
The last example considered natural vibrations of SWNT and BSNT. The natural
frequencies and mode shapes were computed from the equation of motion:
(18.36)
where the mass matrix of the carbon nanotube, [M], was determined from the
distributed masses of carbon atoms and the stiffness matrix, [K], was computed from
Equation 18.20. The stiffness matrix was linearized assuming a small displacement.
The rst several mode shapes of both BSNT and SWNT are plotted in Fig. 18.24
as both ends of the tubes were constrained. The gure shows both axial and bending
vibrations. The rst ve natural frequencies are listed in Table 18.2. Both BSNT
and SWNT have almost the same natural frequencies because the inner cap of BSNT
has little effect on the axial and bending mode shapes. The natural frequencies of
BSNT were generally higher than those of SWNT.
TABLE 18.2
Comparison of First Five Lowest Natural Frequencies of SWCNT
and BSWCNT
1st 2nd 3rd 4th 5th
SWNT (THz) 4.34 5.66 8.21 8.66 9.38
BSNT (THz) 3.99 6.57 8.71 9.72 9.75
M u K u [ ] +[ ] 0
FIGURE 18.24 Comparison of vibrational mode shapes of BSNT and SWNT models.
BSNT SWNT
14
12
10
8
6
4
2
0
0
2
5 5 5 0
14
12
10
8
6
4
2
0
0
2
5 5 5 0
(a)
1st accordion mode shape
BSNT SWNT
14
12
10
8
6
4
2
0
0
2
5 5 5 0
14
12
10
8
6
4
2
0
0
2
5 5 5 0
(b)
2nd accordion mode shape
BSNT SWNT
14
12
10
8
6
4
2
0
0
2
5 5 5 0
14
12
10
8
6
4
2
0
0
2
5 5 5 0
(c)
3rd accordion mode shape
FIGURE 18.24 (Continued)
BSNT SWNT
14
12
10
8
6
4
2
0
0
2
5 5 5 0
14
12
10
8
6
4
2
0
0
2
5 5 5 0
(d)
1st bending mode shape
BSNT SWNT
14
12
10
8
6
4
2
0
0
2
5 5 5 0
14
12
10
8
6
4
2
0
0
2
5 5 5 0
(e)
2nd bending mode shape
BSNT SWNT
14
12
10
8
6
4
2
0
0
2
5 5 5 0
14
12
10
8
6
4
2
0
0
2
5 5 5 0
(f)
3rd bending mode shape
18.6 SUMMARY
In this chapter the concepts pertaining to the static atomic model, coupling
between the discrete atomic model and the continuous nite element model, and
fatigue analysis at the atomic level, and carbon nanotubes were explored. The
static atomic model addressed methodology to predict the static properties of
atomic structures such as Youngs modulus. Then coupling between the atomic
model and a continuous medium was presented for multiscale analysis ranging
from the nanoscale to the macroscale. The fatigue analysis at the atomic level
provides understanding of the fatigue damage process at the nanolevel, leading
to eventual prediction of the fatigue life cycle of the material and to the design
of tougher materials.
PROBLEMS
1. What is the difference between the ab initio simulation method and the
classical molecular dynamics method as applied to nanomechanics?
2. Discuss briey the embedded atom method.
3. Explain clearly how the atomic and nite element models are coupled
together.
4. How is fatigue analysis at the atomic level performed?
5. What are the advantages of performing fatigue analysis at the atomic level
as opposed to conventional fatigue analysis?
6. Compare different atomic energy potentials in terms of their characteris-
tics and applications.
7. Construct the matrix equation of motion for a carbon nanotube with chiral
vector (n, n) or (2n, 0) where n is an integer.
8. Apply the central difference technique for time integration to the matrix
equation of motion derived in Problem 7.
9. Apply Newmarks technique for time integration to the matrix equation
of motion derived in Problem 7.
10. Determine the natural frequencies and mode shapes of the nanotubes
discussed in Problem 7 with appropriate assumptions of the length of the
tube, end closures, and constrained conditions.
11. Model paired dislocation pileups at a free surface and apply a cyclic load
to reveal the movements of the dislocations.
ACKNOWLEDGMENTS
The author greatly appreciates contributions of his former graduate students, R.A.
Duff, S.H. Jung, W.S. Lunt, J.J. Oh., and C. Manthena, in preparing the manuscript.
REFERENCES
1. Kwon, Y.W., Discrete atomic and sheared continuum modeling for static analysis,
Engin. Computations, 20, 8, 964978, 2003.
2. Lennard-Jones, J.E., The determination of molecular elds. I. From the variation of
the viscosity of a gas with temperature, Proc. Roy. Soc. (London), 106A, 441462,
1924.
3. Daw, M.S. and Baskes, M.I., Embedded atom method: derivation and application to
impurities, surfaces, and other defects in metals. Phys. Rev. B, 29, 12, 64436453,
1984.
4. Daw, M.S., Model of metallic cohesion: the embedded-atom method. Phys. Rev. B,
39, 11, 74417452, 1989.
5. Hoagland, R.G., Daw, M.S., Foiles, S.M., and Baskes, M.I., An atomic model of
crack tip deformation in aluminium using an embedded atom potential. J. Mater.
Res., 5, 313324, 1990.
6. Brenner, D.W., Empirical potential for hydrocarbons for use in simulating the chem-
ical vapor deposition of diamond lms, Phys. Rev. B, 42, 15, 94589471, 1990.
7. Felbeck, D.K. and Atkins, A.G., Strength and Fracture of Engineering Solids, Prentice-
Hall, Englewood Cliffs, NJ, 1984.
8. Kahn, H., Ballarini, R., Bellante, J., and Heuer, A.H., Fatigue failure in polysilicon
not due to simple stress corrosion cracking, Science, 298, 8, 12151218, 2002.
9. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-uniform Gases,
Cambridge University Press, London, 1961.
10. Krishnan, A., Dujardin, E., Ebbesen, T.W., Yianilos, P.N., and Treacy, M.M.J., Youngs
modulus of single-walled nanotubes, Phys. Rev. B, 58, 1401314019, 1998.
11. Srivastava, D., Wei, C., and Cho, K., Nanomechanics of carbon nanotubes and
composites, Appl. Mech. Rev., 56, 2, 215230, 2003.
12. Lu, J.P., Elastic properties of carbon nanotubes and nanoropes, Phys. Rev. Lett., 79,
12971300, 1997.
Continuum and
Atomistic Modeling
of Thin Films Subjected
to Nanoindentation
J. David Schall, Donald W. Brenner,
Ajit D. Kelkar, and Rahul Gupta
CONTENTS
19.1 Introduction
19.1.1 Finite Element Model
19.1.2 Need for Atomistic Model
19.2 Modeling of Nanoindentation
19.2.1 Molecular Dynamics Simulation
19.2.2 Embedded Atom Method (EAM)
19.2.3 PARADYN
19.2.4 NordsieckGear PredictorCorrector Algorithm
19.2.5 Temperature Control
19.3 Molecular Dynamics Simulation of Nanoindentation
19.3.1 Simulation Set-Up
19.3.2 Repulsive Potential Indentation
19.3.3 Rigid Indentation
19.3.4 Tests and Comparisons of Indenter Functions
19.4 Conclusions
Problems
References
19.1 INTRODUCTION
Nanocomposite thin lms (Fig. 19.1) usually involve ceramic or polymeric matrices.
The large surface area-to-volume ratio has found large-scale applications in diverse
elds. These materials exhibit improved hardness and remarkable magnetoresistance.
They also exhibit improved properties compared to the parent materials.
1,2
There is
19
also the possibility of new properties not found in the parent constituent materials.
Some of the characteristics of nanocomposite thin lms are as follows:
Superior mechanical properties such as strength, hardness, modulus, and frac-
ture toughness
Important applications of the large surface-to-volume ratio in heterogeneous
catalysis, heat exchangers, and magnetic devices
Thermal stability and heat distortion temperature
Flame retardancy and reduced smoke emissions
Improved chemical and wear resistance
High levels of magnetoresistance and electrical conductivity
To understand the deformation and damage mechanics of these thin lms,
several techniques are being developed and applied. The properties of thin lms
can be different from those of bulk materials even when the chemical composition
of the thin lm and the bulk material are identical. There are several methods to
characterize the mechanical behavior of thin lms. One of the popular methods is
predicting mechanical properties using depth sensing indentation. Although it is
possible to model depth sensing indentation of thin lms using the conventional
nite element method based on a continuum approach, for very thin lms, substrate
effects can be signicant and the conventional continuum principles may not be
applicable. This chapter will address both continuum and atomistic modeling of
thin lms subjected to nanoindentation.
The rst major contribution in indentation experiments was made by Sneddon.
3
Sneddon studied the load, displacement, and contact area relationships for many
simple punch geometries and presented an equation, P = h
m
, where P is the indenter
load, h is the elastic displacement of the indenter, and and m are constants.
4
Values
of the exponent m for some common punch geometries are m = 1 for at cylinders,
m = 2 for cones, m = 1.5 for spheres in the limit of small displacements, and m =
1.5 for paraboloids of revolution.
4
When plasticity is considered while modeling indentation, the constituent
equations become nonlinear and additional material parameters such as yield
strength and work hardening coefcient must be included. Analytical solutions
FIGURE 19.1 Schematic of a thin-lm nanocomposite.
Al
2
O
3
matrix
Ni nano particle
Substrate
for these nonlinear equations are difcult to obtain. Therefore the problems
involving plasticity are studied through experimentation and nite element
simulation.
4
Early indentation studies were done on metals using hardened spherical indent-
ers by Tabor.
5
Later Stillwell and Tabor examined the behavior of conical indenters.
They observed that at least in metals, the impression formed by a spherical indenter
is still spherical with a slightly larger radius than the indenter, and the impression
formed by a conical indenter is still conical with a large included tip angle. Tabor
showed that the shape of the entire unloading curve and the total amount of
recovered displacement could be accurately related to the elastic modulus and the
size of the impression for both spherical and conical indenters.
5
Another important
observation made during these studies was that the diameter of the contact impres-
sion in the surface formed by conical indenters does not recover during unloading
and only the depth recovers.
The indenter must be loaded and unloaded a few times before the load displace-
ment behavior becomes perfectly reversible. The plasticity can be dealt with by
taking into account the shape of the perturbed surface in the analysis of the elastic
unloading curve. Therefore the shape of the unloading curve and the recovered
displacement characterizes the elastic modulus of the material.
Typical load-displacement data obtained from a nanoindenter XP is shown in
Fig. 19.2. This load-displacement data is further analyzed for hardness and modulus
measurements. Stiffness, S, is calculated from the unloading portion of the load
displacement data as follows:
(19.1)
FIGURE 19.2 Schematic of nanoindentation load-displacement behavior.
4
S
dP
dh
E A
r
= =
2
Nanoindentation: load - displacement relationship

P
max
h
max
h
f
S = dP/dh
Loading
Unloading
Indentation depth, h
L
o
a
d
,

P
S = dP/dh is the experimentally measured stiffness of the upper portion of the
unloading data, A is the projected area of the elastic contact, and E
r
is the reduced
modulus (accounting for the effect of nonrigid indenters) dened by the equation,
(19.2)
where E and are Youngs modulus and Poissons ratio for the specimen and E
i
and
i
are the same parameters for the indenter. By measuring initial unloading
stiffness, the modulus is calculated using Equation 19.1 and Equation 19.2.
Equation 19.1 originated from elastic contact theory and was used initially only for
conical indenters. Later it was shown that the equation holds well for spherical and
cylindrical indenters also. Oliver and Pharr
4
proved that Equation 19.1 applies to
any indenter that can be described as a body of revolution of a smooth function.
Finite element simulations by King
6
proved that Equation 19.1 works well for at-
ended punches and triangular cross-sections that cannot be described as bodies of
revolution, with deviations of 1.2% and 3.4%, respectively. King found out that for
all the three geometries, the unloading stiffness can be rewritten as
(19.3)
where is a constant and the values of for different indenter geometries are given
in Table 19.1.
Keeping in mind the difculties and time required for imaging very small inden-
tations, the need for some means other than direct observation of the hardness impres-
sions was felt. Oliver, Hutchings, and Pethica
7
suggested a method based on measured
indentation load-displacement data and an area function (or shape function) i.e., the
cross-sectional area of the indenter as a function of the distance from its tip. This
notion assumes that at the peak load, the material conforms to the shape of the indenter
to some depth.
Doerner and Nix
8
devised a comprehensive method for determining hardness
and modulus from indentation load-displacement data based on the observation that
during the initial stages of unloading, the elastic behavior of the indentation contact
TABLE 19.1
Constant Value for Different Indenter Geometries
Indenter Geometry Value of
Circular 1.000
Triangular 1.034
Square 1.012
Berkovich 1.034
1 1
1
2
2
E E E
r
i
i
=

+

( ) ( )

S
dP
dh
E A
r
= =
2
is similar to that of a at cylindrical punch. Thus the area of contact remains constant
during unloading; i.e., the initial portions of the unloading curves are linear. The
unloading stiffness dP/dh is related to the modulus and contact area through
Equation 19.1. Doerner and Nix proposed a simple empirical method based on
extrapolation of the initial linear portion of the unloading curve to zero load
(Fig. 19.2) and using the extrapolated depth with the indenter shape function to
determine the contact area.
8
Doerner and Nix assumed that the unloading stiffness could be computed from
a linear t of the upper one-third of the unloading curve. They suggested that the
plastic depth h
c
, also termed contact depth (Fig. 19.3), is the distance the indenter
is in contact with the plastically deformed material along the indenter axis during
most of the unloading cycle.
9
The contact area remains constant during the unloading. Once the contact area
is determined for a known indenter shape, using Equation 19.6, the modulus is
calculated from Equation 19.1 and the hardness is calculated according to its
denition:
(19.4)
where P
max
is the peak indentation load and A is the projected area of the impression.
The Doerner and Nix method is based on their experimental observations of metals,
i.e., that the unloading portion of load-displacement curve is linear.
Oliver and Pharr conducted depth-sensing indentation tests on a large number
of materials and suggested that the unloading portion of the load-displacement curve
is not always linear, even in the initial portions of unloading.
4,10
This is particularly
evident while working with hard materials such as ceramics and glass that exhibit
a continuous decrease in indentation depth during unloading due to elastic recovery
FIGURE 19.3 Sectional view of indentation with various quantities used in the analysis.
Indenter
Surface profile after
load removal
P
a
h
Initial surface
Surface profile under load
h
s
h
c
h
f
H
P
A
=
max
within the indentation.
9
Oliver and Pharr reasoned that unloading data are well
described by a simple power law relation given by
(19.5)
where , mand h
f
are constants and are determined by a least square tting procedure.
Oliver and Pharr modied the Doerner and Nix analysis by taking the elastic
recovery into account. The exponent, m, ranges from 1.2 to 1.6.
4
The depth, h
f
, is
the nal depth at which the indenter is in contact with the material (Fig. 19.3). The
Oliver and Pharr method accounts for the nonlinear portion of the unloading curve.
This method provides a procedure for determining the contact depth that should be
used in the area function to determine the contact area at peak load and thus the
hardness. The initial unloading slope is found by differentiating Equation 19.4 at
the peak load and displacement. The contact area A is a function of the contact depth
h
c
given by
(19.6)
The rst term in Equation 19.6 describes a perfect Berkovich indenter and the other
terms describe deviations from Berkovich geometry due to blunting at the tip. Since
the trailing coefcients, begin to approach unity, convergence is achieved by
appropriate decay of the C
n
for large n.
The contact depth h
c
at maximum load, P
max
, is given by
(19.7)
where = 1 for at punch and = 0.72 for conical indenter (Fig. 19.2). For the
paraboloid of revolution, = 0.75.
3,10
Thus, similar to the Doerner and Nix method,
the modulus is obtained from Equation 19.2 and Equation 19.3, and hardness from
Equation 19.4.
The Doerner and Nix method, which considers the at punch assumption, though
simple, underestimates the distance the indenter deects the plastically deformed
material along the indenter axis.
9
Unless the tip is a at punch, the contact area
decreases substantially upon unloading due to elastic recovery within the indentation,
which leads to < 1 in Equation 19.7. The Oliver and Pharr method accounts for
elastic recovery within the indentation and is therefore more accurate than the
Doerner and Nix method. The Doerner and Nix method is a special case of the Oliver
and Pharr method when m = 1. The Oliver and Pharr method initially assumed the
unloading to be elastic. It may be affected by time-dependent deformations such as
plasticity and thermal drift. Plasticity has not been observed in hard materials and
performing the tests in a thermally controlled environment can control thermal drift.
In the present investigation, the elastic properties of multiple-layer aluminum
nitride (AlN) and titanium nitride (TiN) thin lms of various periodicities deposited
P h h
f
m
= ( )
A h h C h C h C h C h
c C c c c c
( ) . = + + + + + + 24 5
2
1
1
2
1
2
3
1
4
4
1
8
CC h
c 8
1
128
h
c
1 128 /
,
h h h h
P
S
c s
= =
max max
max

on silicon (111) substrates are considered. These lms were grown using the PLD
method as abrupt immiscible multiple layers of AlN and TiN, as well as single layers
of the AlN and TiN components.
Conventional nite element analysis (FEA), which is based upon a continuum
description of solids, has been used by several investigators for analyzing indenta-
tion. FEA provides insight into the contact area, plasticity, and effects of a differ-
entiated material substrate on the load-displacement behavior of a thin lm. FEA
requires that the indented body act as a continuum. In the present investigation, an
initial three-dimensional FEA model based on the experiment was 20 mm 20 mm
2 mm, with initial element edge size set at 0.1 mm. The model consisted of 44,000
eight-node three-dimensional (3-D) brick elements and had 48,492 nodes. The rel-
evant details of the nite element model are discussed in the following section.
19.1.1 FINITE ELEMENT MODEL
A number of nanoindentation experiments were performed on thin lm (AlN/TiN)
with silicon substrate. Since experimental results were available for this thin-lm
system, a nite element model was developed for the same system. The mechanical
properties of the thin lm were not known, and it was decided to use the properties
as obtained from the nanoindentation experiments. In the present study, nite element
modeling following substrate and thin lm properties was used.
1. Substrate (silicon (111)): 20 mm 20 mm 2 mm
E = 185 GPA, = .266
2. Thin lm thickness = 250 nm (multiple layers of AlN/TiN)
E = 215
*
GPA, = .25
*
Thin lm modulus was assumed to be 215 GPa.
Thin lm thickness usually is not uniform and can vary with a slope of 10 to
60 nm/mm as measured along the width of the sample. A load of 100 mN was
applied through the nanoindentor tip. Multiple cases were studied with mesh edge
dimension H descending continuously from the 0.1 mm (1E4M) to the 1.0 micron
(1E-6M) level.
Measured thin lm thicknesses ranged from about 250 to 350 nm. Thin lm
hardness and elastic modulus were measured by a frequency-specic depth-sensing
nanoindentation method, which provided continuous stiffness data throughout the
indentation process from the lm surface up to a depth of 1000 nm. In addition, the
thin lm material was assumed to be initially stress free, which may not have been
the actual case. Both lm and substrate were assumed to be homogeneous and
isotropic with perfect elastic-plastic behavior.
In the present nite element simulations, several mesh sizes were utilized. It
was observed that the mesh size has a signicant effect on the prediction of the
through-the-lm thickness displacements. The mesh size was varied from 1E4M to
1E6M. Figure 19.4 shows the three-dimensional nite element model of the thin
lm with substrate. Figure 19.5 shows the simulated through-the-lm thickness
indentation, whereas Fig. 19.6 shows the experimental indentation for the load of
100 mN. The present FEA included both material and geometric nonlinearity. The
nite element results agreed well with the experimental results. Although the lm
thickness in the present simulations and actual experiments was only 250 nm, the
loading on the thin lm was continued until the through-the-lm thickness displace-
ments far exceeded 250 nm. (Simulations were performed until the through-the-lm
thickness displacements were about 750 nm.)
FIGURE 19.4 Three-dimensional nite element model of thin lm showing 2-mm substrate
with 250-nM thin-lm coating.
FIGURE 19.5 Finite element simulation of nanoindentation process.
AIN/TiN thin
film layer
Si substrate
Displacement
7.8e-07
6.7e-07
5.5e-07
4.4e-07
3.3e-07
2.2e-07
1.1e-07
1e-07
Figure 19.7 and Fig. 19.8 clearly show that the experimental hardness and modulus
values within the thin lm thickness region are about 16 GPa and 225 GPa. The modulus
value is the same as the one used in the nite element simulations. Through-the-lm
thickness indentations beyond the lm thickness result into simulating nanoindentation
experiments on the substrate, and hence the modulus has to be far lower than the thin
lm. Finite element simulations exactly predicted similar behavior.
FIGURE 19.6 Atomic force microscope snapshot of nanoindentation.
FIGURE 19.7 Nanohardness (GPa) versus nanoindentation depth (nM).
0 200 400 600 800 1000
9
10
11
12
13
14
15
16
17
AlN-TiN 10 layer film
deposited at 600C
N
a
n
o
h
a
r
d
n
e
s
s

(
G
P
a
)
Depth of indent (nm)
19.1.2 NEED FOR ATOMISTIC MODEL
The FEA of the nanoidentation process is highly nonlinear because (1) the large nodal
displacement forces the stiffness matrix to reformulate throughout the deformation
analysis, (2) the nonlinear behavior of the material transitions from the elastic regime
to the plastic, and (3) the boundary conditions at the sliding indenter contact with
indented material are continually changing throughout the deformation. Furthermore
as the thin lm thickness gets smaller than 100 nm, substrate effects start coming into
the picture. In addition, the nite element analysis also ignores the residual stresses
in the thin lms, which can signicantly affect the prediction of load displacement
behavior in ultrathin lms. An alternative approach is to perform a full three-
dimensional nonlinear analysis using LSDYNA-3D that would render superior char-
acterization for nanoindentation process as well as material behavior and properties
since LSDYNA-3D is an explicit solver with automatic surface-to-surface contact
capability. It should also characterize material behavior and properties better since
LSDYNA-3D has the ability to dynamically adjust the contact elements and geometry
to account for changing surface contact under load.
The present investigation also revealed that although a continuum mechanics
approach works reasonably well when the lm thickness is 250 nm or thicker, smaller
lm thicknesses give rise to convergence problems for nanoindentation modeling,
which represents a large deformation problem. An alternative approach is to model
indentation using atomistic simulation. With todays supercomputers, quantum
mechanics-based analysis of a micron cube of matter for a microsecond would
require over 10 million years of computer time.
11
Therefore large-scale atomistic
simulations typically use some analytic approximation to interatomic forces. One
FIGURE 19.8 Modulus of elasticity (GPa) versus nanoindentation depth (nM).
0 200 400 600 800 1000
175
180
185
190
195
200
205
210
215
220
225
230
AlN-TiN 10 layer film
deposited at 600C
M
o
d
u
l
u
s

o
f

e
l
a
s
t
i
c
i
t
y
such bonding model typically used for metals, the embedded atom method (EAM),
is discussed in the next section in the context of nanoindentation modeling.
19.2 MODELING OF NANOINDENTATION
Many applications of thin lms and coatings require knowledge of their mechanical
properties, which may be signicantly different from those of bulk materials. Based
on analysis of indentation data discussed earlier, by using the continuum nite
element model or by using Hertzian mechanics, it is possible to obtain both local
elastic moduli and hardness of thin lms. However, thin lms often possess residual
stresses, and it is well established experimentally that load-displacement curves
measured by nanoindentation can be altered by stresses in the sample. Atomistic
simulation is a powerful tool for addressing these challenges.
A number of atomistic studies of nanoindentation have been conducted in the
past. Tadmor and coworkers developed a quasicontinuum method capable of repre-
senting very large system sizes and used this model to predict the resolved shear
stress required for emission of a dislocation dipole.
12
Their work also led to impor-
tant insights into the onset of plastic deformation, dislocation nucleation, and dis-
location image forces due to nanoindentation. However, these calculations were
based on an idealized two-dimensional model, which limited its ability to be com-
pared experimentally. A number of workers have applied the embedded atom method
to model nanoindentations. In 1990, Landman and coworkers studied the inuences
of repeated contact.
13
While limited in size, their work was useful in explaining the
jump-to-contact phenomena observed in experiments. More recently indentation
simulations by workers such as Kelchner,
14
Zimmerman,
15
Rodriguez,
16
and
Lilleodden
17,18
have achieved length scales approaching that of earlier experiments.
These simulations used the embedded atom method along with an energy minimi-
zation scheme to explore the issues of onset of plasticity, dislocation nucleation, and
various structural effects such as surface steps and grain boundaries. Zimmerman
and Kelchner
15
also developed a parameterization that allows dislocations and other
defects in the simulated material to be easily visualized and characterized.
The remainder of this chapter will be dedicated to a general discussion of
molecular dynamics simulation, the embedded atom method along with a brief
description of the computational method and parallel algorithm implemented in one
commonly used embedded atom method code, and a description of the typical
temperature control method used in many molecular dynamics simulations. Details
of one commonly used set-up for molecular dynamics simulation of nanoidentation
are also presented and include descriptions of the typical system conguration, two
types of virtual indenters, and results of tests of these indenters.
19.2.1 MOLECULAR DYNAMICS SIMULATION
Molecular dynamics are used to move atoms using classical equations of motion accord-
ing to a model for the interatomic forces. Simulations of this type can be used to determine
equilibrium (and minimum energy) structures or explore nonequilibrium dynamics.
Atomic-scale dynamics are simulated by numerically integrating the classical equation
of motion, i.e., Newtons equation, for each atom:
F = ma (19.8)
(19.9)
where U is the potential energy of the system, r and m are the atomic positions and
masses of each atom, respectively, and t is time.
19
The potential energy U is given
by an analytic expression that yields energy as a function of the relative positions
of the atoms. We present one such energy function known as the embedded atom
method; however, a wide range other of potential functions exist for a wide range
of materials.
19.2.2 EMBEDDED ATOM METHOD (EAM)
The selection of the EAM for the energy function in molecular dynamics simulations
is a popular choice for close-packed metals.
2022
It combines the computational
simplicity required to model large systems with a physical picture that includes
many-body effects and the moderation of bond strengths by local coordination. The
research community has investigated many problems of interest using the EAM,
including point defects, melting, alloying, grain boundary structure and energy,
dislocations, segregation, fracture, nanoindentation, surface structure, and epitaxial
growth.
22
In general, most of the EAM calculations have been carried out in close
connection with experimental work.
The EAM is based on effective medium theory. In this theory, each atom in
a solid is viewed as an impurity embedded in a host material, and the energy of
the atom is a function of the electron density. In the EAM, this density is taken
as the contribution of electron density from neighboring atoms at the site at which
the atom resides. More rigorous theories take this electron density as a sum of the
density in some region surrounding an atom that is determined such that the total
electron density of the solid is included in the calculation. In EAM, however, the
relationship between the site electron density and the potential energy of the atom,
together with a pair term that mimics corecore repulsion, is empirically t to a
range of solid properties. This makes the EAM computationally attractive, as it is
relatively short-ranged and can be computed by summing over pair-wise interac-
tions, while at the same time it is able to reasonably reproduce a wide range of
bulk and defect properties, including properties not used in tting the analytic
form of the potential.
In the EAM formalism, the total potential energy U is taken as a sum of energies
associated with each atom i given by
(19.10)
=
dU
dr
m
d r
dt
2
2
U r F r
i ij
j
embed ij
j
= +

( ) ( )
where (r
ij
) is a corecore pair repulsion of atoms i and j at a radial distance of r
ij
,
and F is an embedding function dened as the energy required to embed atom i in
the electron density (r
ij
). This density is determined through a linear superposition
of the surrounding j atoms. The and F embedding functions are t to various
experimentally determined quantities such as equilibrium density, sublimation
energy, elastic constants, and vacancy formation energy. The corecore repulsion
term (r
ij
) is based on a Coulombic potential with a parameterization derived from
s and p electron densities, as determined by HarteeFock wave functions.
19.2.3 PARADYN
PARADYN, a parallel implementation of Daws EAM code DYNAMO, is used
for the MD simulations. PARADYN uses an algorithm developed by Plimpton
23
that utilizes force-domain decomposition to achieve N/P scaling of computational
time where N and P are the numbers of particles and processors. Each processor
is assigned N/P atoms. The force for all atoms can be described as an N N force
matrix, illustrated in Fig. 19.9, where element ij is the , , or force interaction
between atoms i and j. The total force on atom i is the sum of all elements in row
i of the matrix. The matrix is sparse due to the short range nature of the interatomic
forces and symmetric due to Newtons Third Law. The algorithm for the parallel
implementation of the EAM energy and forces is outlined as follows for the system
illustrated in Fig.19.9
1. Get N/P positions of atoms in their rows. (Processor 4 communicates with
Processor 3.)
2. Transpose its N/P positions with appropriate processor. (Processor 3 trades
with Processor 2.)
3. Acquire N/P positions of atoms in their columns. (Processor 4 commu-
nicates with Processor 2.)
4. Calculate electron density matrix elements in each block.
FIGURE 19.9 Processor assignment of the N N force matrix. Each processor need only store
two pieces of the position and embedding function vectors, each of length N/P to compute
the matrix elements. Processors then communicate this information along rows and columns.
x, F
P = 1 P = 2
P = 3 P = 4
5. Sum values across row. (Processor 4 sums with Processor 3.)
6. Compute embedding functions (F and F ) for each block of N/P atoms
using summed density values.
7. Repeat the rst three steps with F instead of coordinates to get all the F
values for the N/P atoms in rows and columns of the matrix.
8. Calculate the force vectors in each block using F.
9. Sum forces across rows and update positions.
These steps are easy to perform on a parallel machine. Communication takes place
between small groups of P processors and requires information exchange of length
N/P or shorter (versus length N for atom decomposition). Newtons Third Law
can be used to halve the required communication. This is achieved by checker-
boarding the force matrix. This way each pair interaction is calculated only once
in the matrix.
19.2.4 NORDSIECKGEAR
PREDICTORCORRECTOR ALGORITHM
Given the molecular positions and their successive derivatives (velocity, accelera-
tion, etc. at time t), molecular dynamics attempts to determine the positions,
velocities, etc. at time t + t to some degree of accuracy.
29
By assuming that the
classical trajectory is continuous and that no forces act on the particles, the posi-
tions, velocities, accelerations, etc. at t + t can be estimated by Taylor series
expansion.
(19.11)
The superscript p indicates predicted values of position (r), velocity (v), acceleration
(a), and higher derivatives. Unfortunately Equation 19.11 will not predict the correct
trajectories because the forces at each time step have not been taken into account.
To get the corrected positions, etc., the new positions, r
p
, are used to calculate the
force using the rst derivative of the potential energy (i.e., Equation 19.10) and
hence the correct acceleration (a
c
(t + t) = F/m) at time t + t using the embedded
atom formalism (or in principle any other formalism relating force to positional
information). These values are compared to the predicted accelerations to estimate
the error of the prediction step:
a(t + t) = a
c
(t + t) a
p
(t + t) (19.12)
r r v a b
p
t t t t t t t t t ( ) ( ) ( ) / ( ) / ( ) + = + + + + 1 2 1 6
2 3
v v a b
a
p
p
t t t t t t t
t t
( ) ( ) ( ) / ( )
( )
+ = + + +
+

1 2
2
== + +
+ =
a b
b
( ) ( )
( )
t t t
t t
p

The error and results from the predictor step, Equation 19.11 and Equation 19.12,
enter into the corrector step as follows
(19.13)
The values of c
0
, c
1
, c
2
, . . . are chosen to optimize numerical stability and accuracy.
These values for the NordsieckGear
24,25
second-order predictorcorrector using the
rst three derivatives of position are given in Table 19.2. While it is possible to
iterate the predictorcorrector algorithm to achieve higher accuracy, the force cal-
culation is the most time-consuming part of simulation. Thus a large number of
predictorcorrector steps would be inefcient.
19.2.5 TEMPERATURE CONTROL
In the majority of the work presented, temperature control is achieved via the
generalized Langevin equation (GLE).
26
Based on the theory of Brownian
dynamics,
27
this technique assumes that the energy of single particles is exchanged
with the remainder of the system in two ways. The rst is friction, which models
the dissipation of energy into the bulk. The second is by a uctuating force, which
mimics the thermal vibrations of the surrounding atoms. The two balance to produce
the desired average. The equation of motion, which describes these processes, is
given by
(19.14)
where m and v are the mass and velocity of the particle, is a generalized friction
coefcient, and F is a random force. The random forces used in the generalized
Langevin equation have a Gaussian distribution centered about zero given by
(19.15)
TABLE 19.2
Coefcient Values for a Second-Order NordsieckGear
24,25
PredictorCorrector Using Four Values
c
0
= 1/6 c
1
= 5/6 c
2
= 1 c
3
= 1/3
r r a
v v
c p
c p
t t t t c t t
t t t
( ) ( ) ( )
( ) (
+ = + + +
+ = +

0
tt c t t
t t t t c t t
c p
) ( )
( ) ( ) ( )
+ +
+ = + + +
1
2
a
a a a
b

cc
t t ( ) + =
m
d
dt
m
v
v F = +
F R =
2k T
dt
B
Gauss
where is a constant derived from the Debye frequency, k
B
is Boltzmanns constant,
T is the desired temperature, dt is the time step size, and R
Gauss
is a random number
with a Gaussian distribution given by
(19.16)
where R
1
and R
2
are randomly generated numbers between 0 and 1. In PARADYN
the opposing frictional term v and the random force F are added to the total force
before the correction step in the integrator.
19.3 MOLECULAR DYNAMICS SIMULATION
OF NANOINDENTATION
The following sections give an overview of the techniques commonly used to conduct
molecular dynamics simulation of nanoindentation. The rst section describes an exam-
ple simulation set-up and lists parameters for the simulated material presented in this
example, in this case, EAM gold. The next two sections discuss a repulsive potential
and rigid indenter. The last section presents a series of tests of the indenters in EAM gold.
19.3.1 SIMULATION SET-UP
As described previously, the molecular dynamics code PARADYN developed by
Plimpton
26
was used to model the indentation of gold surfaces. PARADYN utilizes
embedded-atom method potentials such as those developed by Foiles, Daw, and
Baskes
21
to model atomistic behavior of close-packed metals. For such materials, the
embedded atom method has proven to be particularly good at modeling bulk and
defect properties such as energies and structures and has been used extensively to
model the onset of plasticity during the nanoindentation of gold in quasistatic
(0 temperature) simulations and to a smaller extent with molecular dynamics simu-
lations at nonzero temperatures. The materials constants derived from the Au potential
used in this study are given in Table 19.3. A crystal lattice was generated by tessel-
lating a unit cell with the desired crystallographic orientation by a desired number
TABLE 19.3
Materials Properties of Gold Derived from the Embedded
Atom Potential: Equilibrium Lattice Constant, Bulk Modulus,
Cohesive Energy, Vacancy Formation Energy, and Elastic
Constants
a
o
4.08 C
11
1.83 nN/
2
B 1.67 nN/
2
C
12
1.59 nN/
2
E
coh
3.93 eV C
44
0.45 nN/
2
E
v
f
1.03 eV
Note: As reported by Foiles, Daw, and Baskes.
21
R R R
Gauss
= 2 2
1 2
log( ) cos( )
of repeat units in the x, y, and z directions. The majority of the simulations conducted
in this work were oriented such that there was a 111 direction normal to the free
surface. A schematic diagram of the model used in the simulated indentation tests
is shown in Fig. 19.10. In the gure, the z direction is taken to be the normal to a
111 plane. The x direction is taken to be along a 110 close-packed direction
and y is along a 112 direction. The bottom layer of atoms is held rigid in the z-
direction and the sides are maintained through periodic boundaries in the x and y
directions.
19.3.2 REPULSIVE POTENTIAL INDENTATION
In the next two sections, two types of so-called virtual indenters will be discussed.
The rst relies on a repulsive potential to generate atomic displacements by the
indenter. The second simply displaces the atoms rigidly in a manner that describes
the geometry of the indenter. In the repulsive potential formulation for the virtual
indenter, the indenter acts as if it were a giant atom with a repulsive force between
the indenter and atoms the indenter contacts. The force from the indenter I on
substrate atom i is given as
(19.17)
where R denes the radius of a virtual spherical indenter, r
iI
is the radial distance
from the center of the indenter to atom i, and n and k are constants related to the
stiffness of the indenter (Fig. 19.11). A strictly repulsive interaction, which is similar
FIGURE 19.10 Schematic diagram of the set-up for molecular dynamics simulation. The x-
axis is along a 110 direction, the z-axis is oriented in the 111 direction, and the y-axis
forms a right-hand coordinate system along a 211 direction (normal to the plane of the
illustration). Periodic boundaries are held in the x and y directions. The bottom layer of atoms
is held rigid. R is the radius of the virtual indenter.
R
z = (111)
x = (110)
Rigid bottom layer
Virtual
indenter
Periodic boundaries
F r R
F k R r r R
i il
i il
n
il
=
= <
0 for
for ( )
to other recent simulations, is intended to mimic nonadhesive contact.
2428
The total
force on atom i is then
F
i
= F
EAM
+ F
iI
(19.18)
where F
EAM
is the total force determined by the embedded atom method and F
iI
is
the force on atom i from the indenter. To simulate loading or unloading of the
indenter, the center of the indenter is placed above the substrate in preparation for
the indention. Typically it is centered in x and y with respect to the periodic bound-
aries of the system. The z component for the center of the indenter is then given as
the maximum z coordinate of the substrate atoms plus the radius of the indenter plus
some small increment of displacement. This initial placement of the indenter ensures
that the indenter is not in contact with the surface at the start of the simulation. To
load or unload the substrate, the z coordinate of the indenter is decreased or increased
by I
depth
/n
steps
at each molecular dynamics time step where n
steps
is the number of time
steps required to indent to a depth of I
depth
. The force on the indenter is then calculated
by summing the forces normal to the surfacefor all atoms in contact with the indenter
(r
iI
< R). The force data and the displacement of the indenter are used to generate
load displacement curves.
19.3.3 RIGID INDENTATION
Placement of the rigid indenter is identical to that of the repulsive potential indenter.
However, instead of adding a force to each atom, a rigid displacement of contacted
atoms is made. At each step the magnitude of the radial vector V
j
between the
coordinate of atom A
j
and the indenter position I(x, y, z) is calculated V
j
= I A
j
.
Contact between atoms and the indenter is dened as follows. For |V
j
| < R where R
is the radius of the indenter, with the atom in contact, the atom is moved along the
radial vector V
j
to the point where | | = R. This generates a rigid displacement
of atoms by the indenter that exactly matches the geometry of the indenter; see Fig.
19.12. The atom is made rigid by zeroing its velocity, acceleration, and higher order
FIGURE 19.11 Schematic diagram of the repulsive potential indenter. A normal force is
applied normal to the surface of the indenter (left panel). The actual displacement of atoms
in contact with the indenter depends on the constants in Equation 19.17.
I(x,y,z) I(x,y,z)
R R
F
r
il
i
V
j
new
derivatives so that the shape of the contact is maintained, i.e., atoms cannot move
back inside the indenter. Once an atom loses contact with the indenter, for instance,
during unloading, it is no longer constrained. For |V
j
| > R, there is no contact. This
type of indenter gives a more accurate estimate of the elastic properties of the
material being indented; however, it works only for loading. During unloading, atoms
lose contact with the indenter and no force is registered.
19.3.4 TESTS AND COMPARISONS OF INDENTER FUNCTIONS
Lilleodden et al.
17,18
have spent considerable effort to examine indenter stiffness on
the virtual indenter function expressed by Equation 19.17. They settled on values of
n and k to be 2 and 10 eV/A
3
, respectively, and utilized a conjugate gradient energy
minimization scheme to determine the equilibrium structure at each indentation dis-
placement before moving to the next. This is equivalent to indenting at a very slow
indentation rate but at zero temperature. Molecular dynamics simulations were per-
formed only at specic points along the load displacement curve where dislocation
emission occurred as indicated by discontinuities in the load displacement curves. We
conducted several small test cases to qualify the indenter function. The results are
signicantly different from those reported by Lilleodden and others where a conjugate
gradient minimization scheme was implemented. This led us to the selection of dif-
ferent values for the indenter stiffness parameters n and k. In our simulations we have
chosen to simulate the entire indentation cycle using molecular dynamics. Due to the
computational limitations inherent to molecular dynamics simulations, namely, the
limitation of time-scale, the trade-off of this choice is a high indentation rate. The
particles in each simulation are held at the thermal equivalent of 300 K via the
generalized Langevin equation thermostat.
26
For a typical indentation simulation of
gold, a molecular dynamics time step of 0.005 ps was used. At each time step the
indenter is moved 0.01 down or up for loading or unloading. This results in an
indentation rate of 2 /ps. Although much faster than experimental indentation rates,
which are on the order of angstroms per second, the indentation rate in our simulations
is still approximately eight times slower than the theoretical dislocation velocity in
FIGURE 19.12 Schematic diagram of the rigid indenter. Atoms are rigidly displaced normal
to the surface of the indenter such that |V
j
| = R.
I(x,y,z)
I(x,y,z)
R R
V
j
A
j
(x,y,z)
j
bulk gold. As a result the underlying plastic deformation mechanisms are largely
independent of the indentation rate with the benet that any thermally activated pro-
cesses present in a typical indentation experiment may be accurately represented.
Illustrated in Fig. 19.13(a) are the load displacement curves of several small-scale
molecular dynamics simulations of indentation with varying values of the constant k
at a constant value n = 1. A 10 radius indenter was used to indent a 111-oriented
substrate with dimensions 70 70 70 to a depth of 5 at a constant
displacement rate of 2 /ps. Small values of k (>0.5 eV/
2
) suggest an overly com-
pliant indenter that leads to little or no plastic deformation of the substrate. Large
values of k (>5 eV/
2
) cause bouncing of the atoms as they make and lose contact
with the tip. This leads to noisy data that makes analysis of the load-displacement
curves difcult. Shown in Fig. 19.13(b) are load displacement curves of the same
substrate with varying values for the exponent n with a constant k = 1 eV/
n+1
. Over
the range of n evaluated no particular value appears to be better than another.
FIGURE 19.13 Load displacement curves for (a) varying values of the constant k with n = 1,
and (b) for varying values of the constant n with k = 1.
60
50
40
30
20
10
0
0
50
40
30
20
10
0
0 2 4 6
2 4 6
F
o
r
c
e

(
n
N
)
F
o
r
c
e

(
n
N
)
Displacement (A)
Displacement (A)
(a)
(b)
n = 1.5
n = 2
k = 10
k = 1
k = 0.1
n = 0.5
n = 1
Based on these tests, we have found suitable values of n and k in Equation 19.17
to be 1.0 and 1.0 eV/
2
, respectively. For the range of indenter stiffness parameters
tested it does not appear possible to back-calculate a value for the modulus or Poissons
ratio of the indenter, a problem also noted by Lilleodden.
18
Therefore all modulus
values reported are for reduced modulus instead of the substrate modulus.
Finite element simulations by several other groups have reported observ-
able periodic boundary effects for substrate thicknesses less than 50 times the
indention depth.
28, 29
To determine the effect of the rigid boundary layer on
indentation results in our simulations, three substrates with thicknesses of 6.3,
12.6, and 19 nm were indented to a maximum depth of 2 nm with a 10 nm
radius indenter. The x and y dimensions were held constant at 30 nm and 30
nm for each substrate indented. The simulated load displacement curves are
shown in Fig. 19.14.
Only minor differences were observed. The load displacement curves for the
19- and 12.6-nm thick substrates overlap completely, while the 6.3-nm thick sub-
strate shows only a slight increase in the force with displacement over the 19- and
12.6-nm substrates. This translates into an increase in hardness of only 2% and a
negligible change in the reduced modulus. These results are consistent with the
nite element simulations of soft thin lms on hard substrates conducted by Chen
and Vlassak.
30
They observed that lm thickness effects have signicant impor-
tance only for indent depths greater than 50% of the lm thickness. Our current
computational resources limit us to simulations consisting of roughly 1,000,000
atoms, which corresponds to the substrate size of approximately 30 nm 30 nm
19 nm, and we limit the indentation depth to 2 nm, roughly 10% of the total
thickness. Additional analysis of atomic stresses during indentation showed that
the in-plane periodic boundary conditions x and y do not inuence the results
signicantly.
FIGURE 19.14 Load displacement curves for substrate thicknesses of 6.3, 12.6, and 19 nm
indented to a depth of 2 nm with a 10-nm radius indenter. The curves overlap completely for
the 12.6- and 19-nm thick substrates with only a slight increase in the force with displacement
for the 6.3-nm substrate with respect to the other substrates.
6000
5000
4000
3000
2000
1000
0
0 5 10 15 20
F
o
r
c
e

(
n
N
)
6.3 nm
19, 12.6 nm
Displacement (A)
Consider the equation for the reduced modulus:
(19.19)
The repulsive potential indenter does not allow accurate calculation of the substrate
modulus, because the indenter modulus E
i
and Poissons ratio
i
are indeterminate.
For the rigid indenter, E
i
goes to innity, and therefore
(19.20)
giving an expression for the reduced modulus that does not include the modulus of
the indenter. The rigid indenter does not allow for unloading, in effect making the
stiffness in the equation
(19.21)
go to innity. The solution lies in the Hertzian analysis. This relation assumes a
linear elastic relation between the indenter and the surface, i.e., no plasticity and no
adhesion. If this requirement is met, the relationship between load and deformation
closely follows the Hertzian model for a rigid, noninteracting parabolic punch
deforming an elastic half-space that can be expressed as a power law expression:
(19.22)
where P is the applied load, R is the radius of the indenter, and d is the depth of
the indentation. If the initial portion of the load displacement curve is t to this
equation, a direct measure of the elastic response of the material is possible. The
simplest way to make the t is by taking the natural log of Equation 19.22.
(19.23)
which is simply the equation for a straight line (y = mx + b). Given the radius R of
the indenter, solving for the reduced modulus E
r
is straightforward. An example load
displacement curve for a 001-oriented substrate is shown in Fig. 19.15(a). The
sample was indented to a depth of 2.5 at a rate of 0.5 /ps with a 40- radius
indenter. No plastic damage was observed. The rigid indenter creates singularities
in the force as individual atoms come in contact with the indenter. It also incorporates
the random frictional forces that are imposed by the GLE thermostat. As a result,
1
1 1
2 2
E E E
r
s
s
i
i
=

( )
+

( )

1
1
2
E E
r
s
s
=

( )
E
S
A
r
=
2
P E Rd
R
=
4
3
3 2 /
ln / ln ln P d E R
r
= +

3 2
4
3
the data appear quite noisy compared to the load displacement curves shown in Fig.
19.13 and Fig. 19.14 for the repulsive potential indenter. To generate good data from
the noise, a running average with a 21-point centered spread is calculated. The results
of this averaging and tting to a LnLn plot are shown in Fig. 19.15(bottom).
The linear t to the data shown in Fig. 19.15 yields a value for the exponential
of 1.52, which is in very good agreement with the Hertzian exponent value of 1.5,
as seen in Equation 19.22. Calculating the modulus with the intercept value of 2.70
and the indenter radius of 40 gives an indentation modulus of 1.77 nN/
2
. This
value is in reasonably good agreement (order of magnitude) with the value of
0.55 nN/
2
for the indentation modulus of bulk EAM gold obtained using the elastic
constants given in Table 19.3. When considering the 12 orders-of-magnitude differ-
ence in the indentation rates between simulation and experiment, this order of mag-
nitude agreement is remarkable. The bulk elastic constants are determined in a static
energy calculation. The use of these constants in the calculation of the indentation
FIGURE 19.15 Load displacement curves for a 001-oriented substrate indented to 2.5 at
a rate of 0.5 /ps with a 40- radius indenter. Top: raw load displacement data,
Bottom: smoothed LnLn data and linear t.
140
120
100
80
60
40
20
20
40
60
0
0.5
0.5 0.5
4.5
3.5
2.5
1.5
0.5
2
1
4
3
2.5 3 1.5 2
0
0
1
1
F
o
r
c
e

(
n
N
)
L
n

(
l
o
a
d
)
1
Ln (displacement)
Displacement (A)
y = 1.5176x + 2.7046
R
2
= 0.9765
0
modulus should in principle only be valid in the case of static loading. This is clearly
not the case in the simulations of nanoindentations presented here. A series of
simulated nanoindentation experiments with indentation rates varying between 2
and 0.01 /ps was conducted on a 001-oriented gold substrate. Even slower
indentation rates (<0.01 /ps) become very computationally expensive. The inden-
tation modulus from these simulations has been plotted as a function of indentation
rate, shown in Fig. 19.16. At slower rates, the indentation modulus appears to
converge toward the static value determined from the bulk elastic constants support-
ing this hypothesis. It is important to note that these approximations are valid only
for shallow indentations where no plasticity is present.
19.4 CONCLUSIONS
Two approaches to modeling indentation have been discussed in this chapter: FEA,
a continuum-based approach, and molecular dynamics simulation, in which the
motion of individual atoms is explicitly followed. For a more comprehensive review
of the literature associated with atomic-level indentation simulations, the reader is
referred to three recent review articles.
31,33
Which method is more appropriate
depends on the conditions to be studied. FEA has the advantage that large structures
and time scales that are comparable can be modeled. A potential disadvantage to
this approach is that the continuum approximation begins to break down as thin
lm sizes become comparable to the spatial extent of defects created during
indentation. On the other hand, fully atomistic simulations are able to describe a
wide variety of defect structures provided that an appropriate interatomic force
model is used, enough atoms can be modeled, and there are enough time steps to
allow nucleation processes (e.g., defect nucleation) to proceed. Enough progress
has been made over the last two decades in the development of quantum-based
FIGURE 19.16 Indentation modulus from simulated indentation of a 001-oriented sub-
strate as a function of the indentation rate. As the rate decreases, the indentation modulus
appears to approach the static value for the indentation modulus (as shown by the dotted line)
determined using the bulk elastic constants.
4.5
3.5
2.5
1.5
0.5
4
3
2
1
0
10 1 0.1 0.01 0.001
Indentation rate (A/ps)
Static indentation modulus 0.55 nN/A^2
I
n
d
e
n
t
a
t
i
o
n

m
o
d
u
l
u
s

(
n
N
/
A
^
2
)
analytic potential functions and in the availability of parallel computing platforms
that the rst two points can be satised for most systems. On the other hand,
numerically integrating classical equations of motion is done in a stepwise fashion,
and generally speaking, faster (rather than more) processors are needed to address
the third point. While progress in this area is being made, it has been much slower
than developments in parallel processing.
Not discussed in the previous paragraph are hybrid continuum atomistic modeling
approaches that are now being explored by a large number of research groups. The
most mature of these is probably the quasicontinuum method developed by Phillips
and coworkers.
34
In this approach, atomic forces are used to govern the deformation
of nite elements in response to external deformations, including indentation. In
related work, Liu
35
has developed a method in which groups of atoms are assigned
to a given nite element region. Where uniform stress is applied, the nite elements
respond, and the atoms in that region follow the continuum deformation. In regions
where the stress distribution has subelement deviations, individual atoms are allowed
to move according to the interatomic forces and any external stresses, and the nite
element containing those atoms distorts in response to the local atomic motion. This
approach allows one to use continuum and atomic dynamics simultaneously in a
manner that can respond to the overall deformation of the system. These and related
multiscale approaches hold great promise for indentation simulations, and with the
rapid advances being made in computing power and numerical algorithms we soon
will be able to routinely model details of indentation across disparate length scales.
PROBLEMS
1. Perform linear nite element simulation of the nanoindentation problem
described in Section 19.1.1 using the same material properties and Berkovich
indentor loading. Generate the force deection curve. Evaluate the effect of
mesh size and perform convergence study.
2. Perform nonlinear nite element simulation of the nanoindentation problem
described in Section 19.1.1 using the same material properties and Berk-
ovich indentor loading. (Include the effects of material and geometry non-
linearity.) Generate the force deection curve. Evaluate the effect of mesh
size and perform convergence study. Compare the linear and nonlinear
analysis results.
3. Explore the effect of thin lm thickness on the nanoindentation results
for thin lm thickness greater than 250 nM as well as lower than 250 nM.
4. Describe the basic differences between the continuum-based nite ele-
ment approach and the EAM-based molecular dynamics modeling
approach for the nanoindentation problem.
REFERENCES
1. Derby, B., Alumina/Silicon carbide nanocomposites by hybrid polymer/pow-
der processing: microstructures and mechanical properties, Solid State Mater.
Sci., 3, 490495, 1998.
2. Sternitzke, M., Review: structural ceramic nanocomposites, J. Eur. Ceramics
Soc., 17, 10611082, 1997.
3. Harding, J.W. and Sneddon, I.N., The elastic stresses produced by the indentation
of the plane of a semi-in nite elastic solid, Proc. Cambridge Philos. Soc., 41,
12, 1945.
4. Oliver, W.C. and Pharr, G.M., An improved technique for determining hardness
and elastic modulus using load and displacement sensing indentation experi-
ments, J. Mater. Res., 7, 6, 15641583, 1992.
5. Tabor, D., Adams, M.J., Biswas, S.K., and Briscoe, B.J., Indentation hardness and
material properties, Solid-Solid Interactions, Imperial College Press, London, 1996.
6. King, R.B., Elastic analysis of some punch problems for a layered medium, Int. J.
Solids Structures, 23, 1657, 1987.
7. Pethica, J.B., Hutchings, R., Oliver, W.C., Hardness measurement at penetration
depths as small as 20 nm, Philos. Mag. A, 48, 593, 1983.
8. Doerner, M.F. and Nix, W.D., A method for interpreting the data from depth-
sensing indentation instruments, J. Mater. Res., 1, 601, 1986.
9. Baker, S.P., The analysis of depth-sensing indentation data, Mater. Res. Soc. Symp.
Proc., 308, 209, 1993.
10. Harding, D.S., Oliver, W.C., and Pharr, G.M., Thin lms: stresses and Mechanical
Properties V, Mater. Res. Soc. Symp. Proc., 356, 663668, 1995.
11. Chung, P.W. and Namburu, R.R., Nanotechnology: Novel Computational Mechanics
Methods for Nanoscale and Nanostructured Solids, Computational and Information
Sciences Directorate, U.S. Army Research Laboratory, Aberdeen Proving Ground,
Maryland.
12. Tadmor, E.B., Miller, R., and Phillips, R., Nanoindentation and incipient plasticity,
J. Mater. Res., 14, 2233, 1999.
13. Landman, U., Atomistic mechanisms and dynamics of adhesion, nanoindentation and
fracture, Science, 248, 454, 1990.
14. Kelchner, C.L. and Plimpton, S.J., Dislocation nucleation and defect structure during
surface indentation, Phys Rev B., 58, 11085, 1998.
15. Zimmerman, J.A., Kelchner, C.L., Klein, P.A., Hamiltion, J.C., and Foiles, S.M.,
Surface step effects on nanoindentation, Phys. Rev. Lett., 87, 165507, 2001.
16. Rodriguez de Fuente, O., Zimmerman, J.A., Gonzalez, M.A., de la Figuera,
J., Hamilton, J.C., Pai, W.W., and Rojo, J.M., Dislocation emission around nano-
indentations on 001 fcc metal surface studied by scanning tunneling microscopy
and atomistic simulations, Phys. Rev. Lett., 88, 36101, 2002.
17. Lilleodden, E.T., Indentation-Induced Plasticity of Thin Metal Films, PhD thesis,
Stanford University, 2001.
18. Lilleodden, E.T., Zimmerman, J.A., Foiles, S.M., and Nix, W.D., Atomistic simula-
tions of elastic deformation and dislocation nucleation during nanoindentation,
J. Mech. Solids, 51, 901, 2003.
19. Allen, M.P. and Tildesley, D.J., Computer Simulations of Liquids. Clarendon Press,
Oxford, 1987.
20. Daw, M.S. and Baskes, M.I., Embedded-atom method: derivation and application to
impurities, surfaces, and other defects in metals, Phys Rev B., 29, 6443, 1984.
21. Foiles, S.M., Baskes, M.I., and Daw, M.S., Embedded atom method functions for the
FCC metals Cu, Ag, Au, Ni, Pd, Pt, and their alloys, Phys. Rev. B., 33, 7983, 1986.
22. Daw, M.S., Foiles, S.M., and Baskes, M.I., The embedded atom method: a review of
theory and applications. Mater. Sci. Reports, 9, 251, 1993.
23. Plimpton, S.J. and Hendrickson, B.A., Parallel molecular dynamics with the embed-
ded atom method, Mater. Res. Soc. Symp. Proc., 291, 37, 1993.
24. Gear, C.W., Numerical Initial Value Problems in Ordinary Differential Equations,
Prentice-Hall, Englewood Cliffs, NJ, 1971.
25. Nordsieck, A., On numerical integration of ordinary differential equations, Math
Comp., 20, 130, 1962.
26. Kubo, R., Fluctuationdissipation theorem, Rep. Prog. Theor. Phys., 33, 425, 1965.
27. McQuarrie, D.A., Statistical Mechanics, Harper and Row, New York, 1976.
28. Bolshakov, A., Oliver, W.C., and Pharr, G.M., J. Mater. Res., 11, 760, 1996.
29. Pharr, G.M. and Bolshakov, A., J. Mater. Res., 17, 2660, 2002.
30. Chen, X. and Vlassak, J.J., J. Mater. Res., 16, 2974, 2001.
31. Harrison, A., Stuart, S.J., and Brenner, D.W., Atomic-scale simulation of tribological
and related phenomena, in Bhushan, B., Ed., CRC Handbook of Micro/Nanotribology,
2nd ed., CRC Press, Boca Raton, FL, 1998, pp. 525596.
32. Brenner, D.W., Shenderova, O.A., Schall, J.D., Areshkin, D.A., Adiga, S., Harrison,
and Stuart, S.J., Contributions of molecular modeling to nanometer-scale science and
technology, in Goddard, W., Brenner, D., Lyshevski, S., and Iafrate, G., Eds., Nano-
science, Engineering and Technology Handbook, CRC Press, Boca Raton, FL, 2002.
33. Schall, J.D., Brenner, D.W., Atomistic simulation of the inuence of pre-existing
stress on the interpretation of nanoindentation data, J. Mater. Res., 19, 31723180,
2004.
34. Phillips, R., Dittrich, M., and Schulten, K., Quasi-continuum representation of
atomic-scale mechanics, Ann. Rev. Mater. Res., 32, 219233, 2002.
35. Zhang, L., Gerstenberger, A., Wang, X.D., and Liu, W.K., Immersed nite element
method, Comp. Meth. Appl. Mechan. Engin., 193, 20512067, 2004.
20
Synthesis, Optimization,
and Characterization
of AlN/TiN Thin Film
Heterostructures
Cindy K. Waters, Sergey Yarmolenko,
Jagannathan Sankar, Sudhir Neralla, and
Ajit D. Kelkar
CONTENTS
20.1 Introduction
20.1.1 Nitrides
20.1.2 Multilayer Coatings
20.1.3 Aluminum Nitride
20.1.3.1 AlN Properties
20.1.3.2 AlN Applications
20.1.4 Titanium Nitride
20.1.4.1 TiN Properties
20.1.4.2 TiN Applications
20.1.5 AlN/TiN Heterostructures
20.1.6 Motivation, Objective, and Organization
20.2 Pulsed Laser Deposition
20.2.1 Laser Energy Influences
20.2.2 Ambient Gas Effects
20.2.3 Substrate Temperature Effects
20.3 Characterization of Thin Films
20.3.1 Thin-Film Thickness
20.3.2 Hardness Testing
20.3.2.1 Testing Basics
20.3.2.2 Microhardness Testing
20.3.2.3 Nanohardness Testing
20.3.2.4 Substrate Effect
20.3.2.5 Indentation Pile-Up
20.3.2.6 Continuous Stiffness Measurement Method
20.3.3 Atomic Force Microscope (AFM)
20.3.4 X-Ray Diffraction (XRD)
20.3.5 Scanning Electron Microscope (SEM)
20.3.6 Transmission Electron Microscope (TEM)
20.4 Performance Evaluation of Thin Films
20.4.1 Approach
20.4.2 AlN Monolayer Film Deposition and Optimization
20.4.3 TiN Monolayer Film Deposition
20.4.4 AlN/TiN Heterostructure Deposition
20.4.4.1 Hardness and Modulus versus Deposition
Temperature
20.4.4.2 Hardness versus Depth of Indentation
20.4.4.3 Hardness versus Layer Characteristics
20.4.4.4 Indentation Pile-Up Considerations and FEM
20.4.5 X-Ray Diffraction
20.4.6 Transmission Electron Microscopy (TEM)
20.5 Optimization of Results
20.5.1 Laser Energy Effect Results
20.5.1.1 Ambient Gas Pressure Effects
20.5.1.2 Location of Deposition on Heater
20.5.2 AlN Monolayer Deposition and Properties
20.5.3 TiN Monolayer Deposition and Properties
20.5.4 AlN/TiN Multilayered Film Properties
20.5.4.1 Hardness and Modulus versus Deposition
Temperature
20.5.5 X-Ray Diffraction (XRD) Data
20.5.6 TEM Data
20.6 Conclusions
Problems
References
20.1 INTRODUCTION
Multilayer coatings and nanothin layers of films are being developed and investi-
gated every day.
18
Previous studies have shown that multilayer coatings can lead
to benefits in performance over comparable single-layer coatings and can combine
the attractive properties of different materials in a single protective layer. There
are many advanced coating systems currently under investigation and many meth-
ods are being researched to create these films, as shown in Fig. 20.1
8
(Elshabini-
Riad). The reason for the enormous amount of research in this area is that these
coatings are useful in a plethora of different fields from national defense, to
biomedical applications, to manufacturing.
2,4,911
Each area of use mandates a
different set of properties and these properties must be analyzed and quantified by
researchers.
20.1.1 NITRIDES
Nitride single-layer coatings have been studied and long used on an assortment
of substrates for their exemplary properties. An interesting note concerning their
mechanical and physical properties is that they behave more as a ceramic than
their constituent metal. A couple of the most popular group III nitrides include
aluminum nitride (AlN) and titanium nitride (TiN). AlN is a wide-band-gap
(energy gap ~6.2 eV) semi-conductor, while TiN is a transition metal nitride thin
film. TiN has the highest melting point and hardness when compared to IV-B
nitrides (Zr and Hf ). AlN also possesses respectable hardness and melting temper-
ature, as shown in Table 20.1
20.1.2 MULTILAYER COATINGS
Researchers have worked on improving the toughness, hardness, and residual stress
state of coated layers in order to create new and enhanced tribological coatings,
often using nanostructural or multilayered coating designs.
12
Multilayer coatings
have many advantages over single-layer films.
15,911,13,14
Multilayering can combine
the attractive properties of single layers into a solitary protective coating.
In addition to a permutation of the best properties, dislocation theory reveals
that the introduction of multiple interfaces parallel to the substrate and potential
crack growth direction will deflect or slow any crack growth. This phenomenon will
in turn increase the toughness and hardness of the coating. The variation in shear
modulus among the materials in superlattice nitride coatings is the primary factor
leading to increased hardness. Thin-film superlattices composed of metal layers,
ceramic layers, and a combination of ceramic and metal layers have been reported
FIGURE 20.1 Classification of the most common deposition processes (Elshabini-Riad
8
).
Thin film
deposition
Chemical
processes
Physical
processes
Sputtering
deposition
Solgel Plating CVD
Evaporation
MOCVD Ion plating RF Electroplating
Laser
ablation
dc PECVD
Thermal MBE Magnetron
Electron
beam
Thermal
to exhibit strength and hardness increases, often >l00%, over homogeneous materials
and rule-of-mixtures values. The barriers to dislocation motion provided by super-
lattice layers with different shear moduli usually account for these results. Koehler
was the first to apply the expressions describing image forces on a dislocation near
a heterointerface to superlattices.
15
While variations of this proposal have been
reported, the strength enhancement is normally predicted to be proportional to the
quotient given by Equation 20.1:
Q = (G
A
G
B
)/(G
A
+ G
B
) (20.1)
where G
A
and G
B
are the shear moduli of the superlattice layers, A and B, respectively.
There are other mechanisms of superlattice strengthening proposed: inhibition of
dislocation motion due to layer coherency strains, increases in total interface area
due to glide across the layers, misfit dislocation arrays at the interfaces, the
HallPetch effect, and the supermodulus effect.
16
Figure 20.2 depicts possible reactions with a layered interface that will influence
hardness and toughness according to the various methods mentioned. There are three
main classifications of nanofilm-layered morphologies.
13
Coating with a limited
number of single layers composes the majority of coatings in use today. These
coatings are useful because they break up the column grain growth and combine
different materials. The second class is composed of coatings with high numbers of
nonisostructural single layers.
The last group is the coating classified as the superlattice. The superlattice is
composed of isostructural single-layer materials that have similar periodic table prop-
erties (i.e., chemical bonding, atomic radii, and lattice dimensions are relatively close).
The thickness of the single layer is similar to its lattice dimension. These superlattices
exhibit increases in hardness and strength when the multilayer superlattice period is
in the range of 5 to 10 nm. Prior work by Mirkarimi et al. with specific nitride
TABLE 20.1
Properties of Various Nitrides
Periodic
Group
IV B V B VI B III A
IVA
Elements Ti Zr Hf V Nb Cr Mo B Al Si
Nitride Properties
Melting
point (K)
3220 3150 3130 2619 2846 2013 2223 3300 2523 2173
Hardness
(bulk, GPa)
21 15 17 15 14 11 17 33 12 17
Crystal
structure
fcc fcc Cubic fcc fcc hcp/fcc Cubic/
hcp
hcp
Dominant
bond type
Metallic Covalent
fcc = face-centered cubic.
hcp = hexagonal close-packed.
superlattice coatings has achieved maximum hardness values for coatings of TiN/VN
(~56 GPa), and of TiN/NbN (~51 GPa) that were much higher than the hardness of
17 to 23 GPa for homogeneous single-crystal coatings of TiN, VN, and NbN.
7,17,18
The use of TiN/AlN multilayered thin-film coatings has proven to be successful.
Research has proved that 0.5-Pm thin films achieve the same performance as mono-
nitride TiN films with more than six times the thickness.
18
Studies involving alternating layers of Ti and TiN multilayer coatings in the
TiN system were performed by Bull et al. to demonstrate the advantages of com-
positionally and structurally modulated coatings over conventional single-layer
titanium nitride coatings in tribological applications.
4
Bull classified two different
multilayer coatings:
1. Structural multilayers The applied substrate bias is cycled between two
values to produce layers with different levels of residual stress. In this
case, bias voltages of 30 V and 60 V were used, which produce com-
pressive stresses in single-layer coatings of ~3 GPa and ~6 GPa, respec-
tively. The multilayer coating should thus have a lower residual stress
than a comparable single-layer coating deposited at 60 V bias.
2. Compositional multilayers The flow of reactive gas to the chamber
(nitrogen in this case) is stopped periodically to produce alternating layers
of titanium and titanium nitride. Owing to the relatively high process
pressures used in SIP and the restrictions in pumping required to maintain
these pressures, it takes a few minutes to completely remove the nitrogen
from the chamber. For this reason, the coatings are expected to have graded
composition across the Ti/TiN interface.
4
FIGURE 20.2 Periodic layered structure and toughening mechanisms (Holleck
3
).
Layer mat. II
Crack
Crack
Crack
Interface
Interface
Inhibition of grain growth (crack splitting,
interface toughening, stress relaxation)
Crack deflection (reduced crack propagation)
Interface opening or delamination (reduced
stress concentration)
Nanoplasticity (energy dissipation, reduced
crack propagation, stress relaxation)
Substrate
Layer mat. I
Layer mat. II
Layer mat. I
An important concept in the production of multilayer coatings is the multilayer
period O, which is the repetitive spacing in the structure. Bull assessed the properties
of the multilayer coatings as a function of O. For structural multilayers, at low O the
hardness was less than that of a comparable single-layer film, but increased up to a
maximum at O~10 nm. The hardness of very thick structural multilayers is indistin-
guishable from the average value of the single-layer coating hardness.
4
The period
thickness in this investigation is varied to address different types of films. Interatomic
distances for AlN (0001) and TiN (111) are used to calculate the lattice mismatch
at the AlN/TiN interface, which is found to be 3.9%.
20.1.3 ALUMINUM NITRIDE
20.1.3.1 AlN Properties
Thin films of AlN are very promising dielectric materials because of their wide band
gap, e.g., 6.2 eV. Generally speaking, a wide band gap semiconductor has an energy
band gap wider than approximately 2 eV. Other examples of wide band gap semi-
conductors are gallium nitride, 3.4 eV, and silicon carbide, between 2.2 and 3.25
eV. AlN maintains a wurtzite crystal structure with lattice constants: a = 3.112
and c = 4.982 , as shown in Fig. 20.3. A is the typical wurtzite crystal structure,
B is the view along the [0001] direction, C is the view along the [1120] direction,
and D is the view along the [0110] direction. Dark atoms are the aluminum atoms
and the lighter atoms are the nitrogen atoms.
AlN thin films grown on silicon are oriented with the c-axis of the hexagonal
lattice perpendicular to the substrate.
19
AlN has been shown to form a nonequilibrium
metastable cubic zinc blende. AlN has also been found to have high thermal con-
ductivity with a thermal expansion coefficient close to that of silicon.
20
This is the
FIGURE 20.3 Various views of AlN wurtzite structure.
reason for its wide use as the initial layer on silicon substrates for many hetero-
structures. Surface tensions can be reduced with a greater match in thermal and
crystallographic properties such as those maintained by AlN.
20.1.3.2 AlN Applications
AlN exhibits a remarkably high thermal conductivity, which is second only to
beryllia. Its thermal conductivity exceeds that of copper at moderate temperatures
(~200C). This high conductivity coupled with high volume resistivity and dielectric
strength leads to its use as a substrate and as packaging for high power or high
density assembly of microelectronic components. Substrates made from AlN provide
more efcient cooling than conventional or other ceramic substrates; hence, they are
used as chip carriers and heat sinks. Other properties of AlN lead to its use in electro-
optic applications. AlN, with its large energy gap, may also be the material of choice
for passivation and the MISFET gate insulator due to its substantial resistivity and
breakdown electric elds.
21
There is a heightened interest in wide-band-gap semi-
conductors for two important classes of applications, namely blue/green light emit-
ters and high power/high temperature electronics.
22
Basic properties of both AlN
and TiN thin lms are listed in Table 20.2.
20.1.4 TITANIUM NITRIDE
20.1.4.1 TiN Properties
Titanium nitride thin crystalline lms usually exist as -TiN with a NaCl-type
lattice.
23
This is an fcc-type structure, as shown in Fig. 20.4. In this gure the dark
atoms are the titanium atoms and the lighter atoms are the nitrogen atoms. The
arrangement of the transition metal nitrides crystal structure is determined by the
proportion of the atomic radius of the interstitial element and the transition metal,
TABLE 20.2
Basic Aluminum Nitride and Titanium Nitride
Thin-Film Properties
Aluminum Nitride Titanium Nitride
Lattice constants () wurtzite a = 3.112; c = 4.982
zinc blende a = 4.38
FCC (NaCl)
a = 4.24
Band gap (eV @ 300 K) wurtzite 6.2
zinc blende 5.11 (theory)
N.A.
Index of refraction (3 eV) 2.15 0.05 N.A.
Electrical resistivity (-cm) 10
13
14 10
6
Thermal expansion
( 10
6
/C)
a/a = 4.2
c/c = 5.3
a/a = 9.4
Thermal conductivity (W/mK) 200 typical
320 pure single crystal
1167
Density 3.257 g/cm
3
NA = not available.
r r
x
/r
Me
, according to the Hagg rule.
24
If r is less than 0.59 (r
N
/r
TiN
0.56), simple
structures such as NaCl or a simple hexagonal will be formed, and if r is larger than
0.59, the transition metal and the interstitial atoms will form very complicated
structures with unit cells that contain as many as 100 atoms.
10
The elastic strain energy for TiN is minimum when the [111] direction lies
normal to the film and substrate.
25
The dominant crystal orientations are those that
exhibit high growth rates because they tend to take over the other directions so that
the trend is growth with a [111] preferred orientation.
These films have the following properties: excellent corrosion and wear resistance,
relative inertness, maintenance of a high sublimation temperature, high thermal and
electrical conductivity, and high hardness. These properties are displayed in
Tables 20.2 and 20.3. The mechanical and electrical properties of titanium nitride thin
films have a significant dependence on the microstructure and therefore on the growth
conditions. As in most known materials, their imperfections play a great role in their
properties. Voids and impurities notably influence the mechanical properties such as
hardness and the electrical properties such as resistivity.
9
In addition, kinetic restric-
tions imposed on the PLD growth process by nonequilibrium processes influence the
microstructure.
TiN is utilized because of these properties and its low friction and admirable
resistance to oxidation at temperatures below 500C. These properties are valuable
when choosing a material for an optical or electronic device.
FIGURE 20.4 Left: crystal structure of titanium nitride, FCC; right: preferred growth orien-
tation for TiN [111].
TABLE 20.3
Mechanical Properties of Aluminum Nitride and Titanium
Nitride Films
Aluminum Nitride Titanium Nitride
Modulus of elasticity (GPa) 260350 350450
Hardness (GPa) 1825 2040
Poissons ratio 0.24 0.25
20.1.4.2 TiN Applications
The growth of ultrathin TiN heterostructures on various substrates is a subject of
intense research because there is an increasing miniaturization drift in microelec-
tronics. Smaller appliance sizes and increased densities require the development of
reliable electronic devices at the nanometer scale.
26,27
Special attention is given to
the morphological evolution of TiN heterostructures.
Known for its temperature stability and mechanical hardness, TiN is currently
utilized in integrated circuits and has been applied in a wide range of industrial
areas, such as hard coatings, coatings for corrosion resistance and decoration, and
diffusion barriers in microelectronic devices.
13,78,2830
To study the dielectric prop-
erties of laser-processed BaTiO
3
thin films on Si (100), researchers have used a
conductive buffer layer of TiN because it has the following advantages: low diffusion
coefficients of metal atoms in the TiN layer, low solubility, and chemical and thermal
stability.
31
TiN is a good contact material for wide-bandgap nitrides, because its
electrical conductivity is high and its contact resistance is low.
TiN coating is frequently selected as a protective film for a variety of metal
surfaces in order to increase service life. The coating of stainless steel is a widely
known application.
32
The corrosion resistance of stainless steel can be increased and
its surface hardness can be improved by coating with a TiN film. This also adds a
golden decorative color.
33
Recently TiN research has focused on its applications in semiconductor device
technology. The packing density of ultralarge integrated circuits is steadily increasing
and this creates a need for corresponding decreases in the thickness of the metal
interconnect layers and diffusion barriers.
34
The early stages of the growth of TiN
thin layers have been studied extensively and there is a great demand for highly
conductive, ultrathin, and ultrasmooth TiN layers.
27
Table 20.3 displays the standard
mechanical thin films properties for AlN and TiN.
20.1.5 AlN/TiN HETEROSTRUCTURES
In many instances the properties of a single material are not sufficient for an
application. This difficulty is overcome by the use of a multilayer coating that merges
the attractive properties of several materials to solve problems in the application.
Simple examples of this include the use of interfacial bonding layers to promote
adhesion and thin inert coatings on top of wear-resistant layers to reduce the corro-
sion of cutting tools.
4
Research now shows that the multilayer structure produced
by the deposition of many alternating layers of two materials improves performance
more than the use of a mixed coating produced by the introduction of new interfaces.
This occurs even when the two materials do not have definite functional requirements
in the intended application.
In order to be used as a cutting tool material, the layers in a material must have
the ability to be employed above 1000C and the AlN/TiN multilayers are useful
for this reason. They are not miscible at high temperatures (>800C) and have a
high oxidation resistance.
36
In Japan, Setoyama et al. have investigated the TiN/AlN polycrystalline superlattice
system. These researchers deposited the films by reactive cathodic arc deposition in
an opposed cathode system and discovered a peak hardness of approximately 39 GPa
at a superlattice period of 2.5 nm.
7
It has been discovered that the initial oxidation
temperature for the TiN/AlN superlattice film is 930C, which is significantly higher
than the value of 790C that has been reported for TIN.
7
TiN/AlN-coated tools
maintain their exemplary wear properties for several reasons. When utilized as cutting
tools at very high speeds, the tool tip temperature can reach 1000l l00C and the tool
must resist oxidation as well as interdiffusion of individual layers in the superlattice
films. If oxidation occurs, the films lose their shielding properties because the oxides
do not have the mechanical strength to withstand the loads caused by the cutting
procedures.
Sproul et al. found a curious result in their research. When TiN/AlN films were
manufactured as a superlattice, the TiN layers in the TiN/AlN superlattice manipu-
lated the structure of the AlN layers. At a superlattice period of 2.5 nm, the AlN
layers assume the same structure as TiN (i.e., the NaCl structure).
7,8
AlN forms a
hexagonal structure during equilibrium, but in this circumstance TiN acted as a
model and forced AlN into the NaCl structure.
In addition to cutting tool advantages, AlN/TiN structures are useful as capacitors
and transistor gate elements of high-temperature electronic circuits. AlN is a covalent
material and therefore does not adhere well to metallic substrates, so it performs
poorly as a single layer. This issue can be resolved when AlN is adhered to covalent
hard materials such as TiN. This creates bonded multilayer composites that can
be used as wear-resistant materials, with significant improvements in the high-
temperature properties.
2
20.1.6 MOTIVATION, OBJECTIVE, AND ORGANIZATION
Thin films offer many advantages over bulk materials and researchers are contin-
ually searching for smaller and more efficient devices. Pulsed laser deposition
(PLD) is a good technique for the growth of many new thin, stoichiometric films.
By changing PLD parameters, the AlN and TiN thin-film properties can be changed
and optimum values can be obtained. Material composition and microstructure
will be adjusted in this research to analyze the optimization of strength and
toughness in the coatings. This AlN/TiN multilayer study will report hardness
gains and losses in the multilayer thin films. The overall mission of this research
is to understand the synthesis, optimization, and characterization of AlN/TiN thin-
film heterostructures.
20.2 PULSED LASER DEPOSITION
The thin films grown for this work were deposited using a vacuum deposition
technique. Figure 20.1 lists other deposition methods, which vary as to how the
deposition flux is generated, the means used to control the arrival rate of the film
species, the energy of the deposition flux, and the pressure in the vacuum chamber.
The PLD method was used for the deposition of films in this research. PLD has
been used as a deposition technique since the 1960s. This process is discussed
completely in Chapter 10.
20.2.1 LASER ENERGY INFLUENCES
The reach and contour of the plume can be controlled by varying the fluence. Laser
fluence is the energy in the area where the laser spot hits the target. It is calculated in
J/cm
2
and can be varied by changing the laser pulse energy or by regulating the laser
spot size with a change in the aperture through which the laser beam exits the laser
chamber. This influences the plumes appearance both in shape and size. The fluence
has a minimum value that corresponds to the threshold fluence of the material. Below
this threshold fluence, the energy is not great enough to create an explosion. A higher
fluence will produce a larger explosion, which will make the plume more round shaped,
but not necessarily longer.
The quick, strong heating of the target surface by the intense laser beam (typi-
cally up to temperatures of more than 5000 K within a few ns, corresponding to a
heating rate of about 1012 K/s) guarantees that all target components, even those
with different partial binding energies, evaporate simultaneously. When the ablation
rate is high enough (as with laser fluences well above the ablation threshold), a
Knudsen layer is formed and further heating creates a high temperature plasma that
then expands adiabatically in a direction perpendicular to the target surface.
Researchers established that the YBa
2
Cu
3
O
7
thin films greatest thickness
increases with an increase in the pulse energy density.
37
A significant part of the
irradiated volume is ejected as a plume consisting of molecular clusters of different
sizes as well as individual molecules. At lower energies, larger clusters are ejected
and a rougher surface remains after the ablation. Below the threshold value, the
amount of ejected material drops sharply and single molecules are primarily
desorbed.
38
A nonlinear increase in the maximum thickness of the deposit is also
found. The maximum thickness increases very rapidly for low energy densities.
However, Fahler found that for stainless steel, laser energy density increases with
the deposition rate.
39
Another study demonstrated that high laser energy density can
produce better quality film surfaces and this may be due to the higher fluence
producing a higher particle density.
40
Studies with diamond-like-carbon (DLC)
coatings established that hardness and modulus increase with an increase in laser
intensity.
41
The energy of the laser will affect several properties including the
film thickness and its uniformity, deposition rate, particulate density, and the
mechanical properties of the film.
39,40
The pulse energy density plays a very important
role in controlling the thickness and composition of PLD thin films. Singh and
Narayan experimentally observed that the deposition area is much more expansive
at lower energy densities. However, it became more forward directed as the energy
density was increased.
37
20.2.2 AMBIENT GAS EFFECTS
Another parameter that affects the plume is the deposition ambient gas pressure.
When the pressure increases, the plume has more gas molecules to collide with and
its reach diminishes. The plume then condenses onto the substrate, which is kept at
a suitable distance from the target. A reactive environment within the vacuum
chamber can be introduced to modify either the growth rate or film composition.
When the PLDs of alumina and zirconia thin films on polymers and glass for optical
and protective coatings were investigated, it was found that the film refractive index
decreased with an increase in gas pressure.
42
Similar experimental results have been
explained by the scattering of a dense cloud of ablated material in a diluted gas.
Sturm studied scattering of Ag in argon gas. Since the mass of an Ar atom is
sufficiently dense for a large scattering angle, approximately 90% of all scattering
angles are larger than 10, as estimated from ballistic collisions (Fig. 20.5).
43
There-
fore, the scattered ions are not deposited onto the substrate. As the colliding Ar
atoms are removed from the flight path, the ensuing slower ions and atoms fly
through a significantly reduced Ar atom density and are more likely to reach the
substrate. Sturm found that the faster ions are scattered and not deposited, and that
they open a channel for the deposition of the slower particles. This model was
independent of the exact values of the cross-section and the material.
Deposition in an inert gas changes the kinetic energy distribution of the deposited
particles and notably lessens the resputtering at 0.04 mbar. The dependence of the
deposition rate on pressure below 0.1 mbar can be explained by the scattering of a
dense cloud of ablated material moving through a diluted gas.
Other studies confirmed this model. Geohegan states that ablation into back-
ground gases results in scattering and attenuation of the laser plume.
44
Research
with carbon films found that the ablated carbon species are decelerated to a degree
that depends on the background pressure when PLD occurs in a gas ambient.
Alexandrous research found that during CN
x
deposition, the carbon species
react effectively with N
2
at the beginning of their trajectories. However, when they
reach the substrate, they usually do not have adequate kinetic energy to form a hard
coating. Additional studies on the effects of ambient helium gas pressure and ambient
argon gas pressure on the optical, structural, and physical properties of carbon (C)
films deposited by pulsed laser ablation (PLD) using a camphoric carbon target have
shown that the surface morphology, structural, and physical properties of deposited
films are strongly dependent on the choice and pressure of the ambient inert gas.
Research performed specifically on nitrides has demonstrated that the nitrogen
content and film quality of SiN
x
films depend on the laser fluence and the N
2
gas pressure.
FIGURE 20.5 Model for dense cloud scattering of ablated material in an Ar gas (Sturm
43
).
Ag
+
, 100 eV
Ag-ions
Ag-atoms
Ar, 0.1 eV
The film properties are determined by reactions between the plume and the N
2
gas,
in particular, by reactions between Si species and nitrogen radicals in the plume and
collisions between SiN
x
clusters and nitrogen molecules.
20.2.3 SUBSTRATE TEMPERATURE EFFECTS
The temperature of the substrate has a considerable effect on the early development and
growth of the film. The homologous temperature, T
h
T
s
/T
m
, where T
s
is the temperature
of the substrate and T
m
is the melting point of the film material in Kelvin, has been
found to affect the overall microstructure. During this study the temperature was varied
from 200qC up to 800C, for example, when work was performed at T
s
600C, T
h
0.27 and 0.34 for TiN and AlN, respectively. A higher substrate temperature increases
the atoms energy, which increases their mobility. When the particles have more energy,
the grain boundary has more mobility, and grains grow in size to lower their strain energy.
The details of film nucleation and growth demonstrate the dependence of film
growth on material parameters and deposition conditions. The temperature at which
the film is grown is a fundamental parameter governing the thin-film microstructure.
The paramount temperature for optimal quality film growth is the regime where there
is adequate surface diffusion to permit surface atoms to minimize their surface energy,
and therefore become thermodynamically stable.
45
Research demonstrates that higher
energy atomic fluxes, on the order of 5 eV to 15 eV, improve a variety of film properties.
Past studies on the ablation of TiN on various substrates have shown that TiN
formation is controlled by the reactivity of the ablated and scattered species formed
by the strong interaction between the ablation plume and gas pulse, which is maintained
by subsequent adiabatic expansion to the substrate.
46
True epitaxial growth, however,
is controlled by surface diffusion, which is also shown to be dependent on the substrate
temperature.
46
The surface temperature of the substrate determines the atoms surface
diffusion ability. High temperature favors rapid and defect-free crystal growth, whereas
low temperature or large supersaturation crystal growth may be overwhelmed by
energetic particle impingement, resulting in disordered or even amorphous structures.
Studies by other researchers have shown that high temperature can reduce film surface
roughness. CdO-doped SiO
2
films were found to have an optimum substrate temper-
ature for quality crystalline films free from the CdO sites between 200C and 300C.
Crystalline size has been shown to increase when the substrate temperature is increased
for V
2
O
5
films. Xiaohua Liu, on the other hand, found that TiO
2
thin-film surface
roughness increased with an increase in substrate temperature.
47
20.3 CHARACTERIZATION OF THIN FILMS
20.3.1 THIN-FILM THICKNESS
Paramount to the measurement of any films mechanical or electrical property is the
measurement of the films thickness t. The thickness of each film created was
measured with an optical interferometric profilometer, the Wyko RST-500, as dis-
played in Fig. 20.6, and a Pacific Nanotechnology Nano-R AFM.
The Wyko RST-500 is a phase-shift interferometry-based metrology instrument.
An optical profilometer has advantages over the traditional stylus profilometer.
Sensitive surfaces can be measured nondestructively and with the use of camera
images. The results are not aggregated linear measurements but contiguous three-
dimensional surface maps. An optical interferometer functions on the principle of
interference light waves that divide a wavefront into two or more parts, a reference
wavefront and a measurement wavefront, which then travel different paths before
combining to produce an interference pattern, also known as an interferogram. A
phase shifting method is used to extract the height information. The intensity is
measured at each point in the interferogram as the phase between the measurement
and reference waves is changed in discrete steps.
The steps were measured for each deposition condition (i.e., temperature, laser
energy, and ambient gas pressure). This step measurement was performed by mask-
ing each sample with a clean-edged silicon strip. A large silicon sample equal to
the size of the heater was mounted and a deposition was run in the initial stage of
the research. These results suggested a much thicker film directly under the plume.
The schematic of the masked sample is displayed in Fig. 20.7 and the resulting
sample after deposition is shown in Fig. 20.8. Several traces from representative
thickness profiles obtained with the Wyko profilometer are displayed in Fig. 20.9.
20.3.2 HARDNESS TESTING
20.3.2.1 Testing Basics
Hardness implies the resistance to local deformation. Depending on the material,
hardness involves different things. For example, hardness is a gauge of the indentation
depth for a plastic material (i.e., a solid material that is indented beyond its yield point).
FIGURE 20.6 Wyko RST-500 optical interferometric profilometer.
For a brittle material, hardness is defined in the course of scratch experiments.
Scratch tests reveal the oldest form of hardness measurements, typically referred to
as Mohs hardness.
48
The quasistatic indentation hardness method by Tabor of more
familiar today.
20.3.2.2 Microhardness Testing
The Vickers hardness test is a standard form of microhardness testing that executes
calculations from the size of an impression produced under load by a pyramid-
shaped diamond indenter, as shown in Fig. 20.10. Created in the 1920s by engineers
at Vickers, Ltd. in the United Kingdom, the diamond pyramid hardness test allows
FIGURE 20.7 Model of sample used to mask the substrate and measure the film thickness.
FIGURE 20.8 Large masked silicon wafer and resulting deposition pattern.
Substrate
heater
Substrate
mask
Si (100)
substrate
Thickness
measured along
the entire length
F
I
G
U
R
E

2
0
.
9
W
y
k
o

f
i
l
m

t
h
i
c
k
n
e
s
s

p
r
o
f
i
l
e

t
a
k
e
n

f
o
r

a
n

A
l
N

t
h
i
n

f
i
l
m
.
1
.
2
0
.
0
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
1
.
2
1
.
6
0
.
2
0
.
4
0
.
6
0
.
8
2
0
0
1
5
0
1
0
0
5
0 0
5
0
R
q
:

5
4
.
5
0

n
m
R
q
:

5
5
.
6
6

n
m
R
a
:

5
5
.
4
6

n
m
R
t
:

1
2
2
.
6
7

n
m
R
p
:

6
5
.
9
8

n
m
R
u
:
5
6
.
7
0

n
m
R
a
:

4
7
.
2
7

n
m
R
t

:

1
8
8
.
4
9

n
m
R
p
:

3
5
.
3
6

n
m
R
u
:
1
5
3
.
1
3

n
m
L
:

0
.
0
0

m
m

6
0
.
6
1

n
m
R
:

1
.
5
6

m
m

-
-
-
D
:

1
.
5
6

m
m

-
-
-
A
n
g
l
e
:

-
-
-
C
u
r
v
e
:

2
9
.
5
2

m
T
e
r
m
s
:

N
o
n
e
A
v
g
.

h
t
:

0
.
2
7

n
m
A
r
e
a
:

0
.
4
2

u
m
2
L
:

0
.
0
0

m
m

1
5
3
.
1
3

n
m
R
:

1
.
1
6

m
m

-
-
-
D
:

1
.
1
6

m
m

-
-
-
A
n
g
l
e
:

-
-
-
C
u
r
v
e
:
4
8
.
2
5

m
T
e
r
m
s
:

N
o
n
e
A
v
g
.

h
t
:

5
6
.
2
7

n
m
A
r
e
a
:
6
5
.
2
0

u
m
2
0
.
0
0
6
0
4
0
2
0 0
2
0
4
0
6
0
8
0 0
.
0
0
0
.
5
0
1
.
0
0
1
.
5
0
0
.
2
0
0
.
4
0
0
.
6
0
0
.
8
0
1
.
0
0
m
m
m
m
n
m n
m
1
.
1
2
3
,

1
1
7
M
o
d
e
:

P
S
I
M
a
g
:

4
0

X
2
D

p
r
o
f
i
l
e
s
X
-
p
r
o
f
i
l
e
/
2

P
t
/
r
a
d
i
a
l
Y
-
p
r
o
f
i
l
e
/
c
i
r
c
u
l
a
r
0
9
/
2
6
/
0
2
1
1
:
5
2
:
2
9
m
m
S
i
z
e
:

3
6
8

2
3
6
T
i
t
l
e
:

A
I
N

s
i
n
g
l
e

l
a
y
e
r
N
o
t
e
:

p
o
s
i
t
i
o
n

1
m
m
C
o
p
y
r
i
g
h
t

2
0
0
6

b
y

T
a
y
l
o
r

&

F
r
a
n
c
i
s
the establishment of a continuous scale of comparable numbers that accurately
reflect the wide range of hardnesses found in steels. The indenter employed in the
Vickers test is a square-based pyramid whose opposite sides meet at the apex at
an angle of 136. The diamond is pressed into the surface of the material at loads
ranging up to approximately 120 kg force, and the size of the impression (usually
no more than 0.5 mm) is measured with the aid of a calibrated microscope. The
Vickers number (HV) is calculated using the formula, HV 1.854(F/D
2
), with F
being the applied load (measured in kilograms force) and D
2
the area of the
indentation (measured in square millimeters). The applied load is usually specified
when HV is cited. The Leco M400-H1 hardness tester is displayed in Fig. 20.11.
20.3.2.3 Nanohardness Testing
Nanoindentation has been proven particularly useful in measuring the hardness
and stiffness of thin films by reducing the hardness size range by one step. The
MTS nanoindentation tester is shown in Fig. 20.12. The indentation response of
the film on a substrate is a complicated function of the plastic and elastic properties
of both the film and the substrate. As the indenter is driven into the material, both
elastic and plastic deformation occurs, which results in the development of a
hardness impression conforming to the contour of the indenter to some contact
depth h
c
.
It is possible to accurately observe both the load and the displacement of the
indenter during indentation experiments in the nanometer range. By means of the
established techniques of Oliver and Pharr
49
or those of Doerner and Nix,
50
the
hardness and elastic modulus can be found from the peak load and the initial slope of
FIGURE 20.10 Schematic of a Vickers hardness indenter tip.
136 between
opposite faces
d
1
d
2
the unloading curves. Both of these established methods depend on estimating the
contact area under the load, which is difficult especially when pile-up occurs.
51
Modeling the indentation to include plasticity is a more complex problem. The
equations are nonlinear and additional material parameters must be included (e.g.,
FIGURE 20.11 Leco microhardness tester.
FIGURE 20.12 MTS nanoindenter XP.
yield strength and work hardening). As a result, plasticity in indenter contact
problems requires a great deal of experimentation and finite element simulation
to be understood.
49
Early experiments with metals by Tabor
48
revealed that the impression created
by a spherical indenter remains spherical with a slightly larger radius than the
indenter, and the impression formed by a conical indenter remains conical with a
larger tip angle.
49
This confirmed that the diameter of the contact impression does
not recover during unloading; only the depth recovers. Samples of various types
of loading situations are displayed in Fig. 20.13. The significance of these experi-
ments is that the plasticity can be dealt with by taking into account the shape of
the perturbed surface in the analysis of the elastic unloading curve.
49
Therefore,
the shape of the unloading curve and the recovered displacement can be related to
the elastic modulus.
The basic theory of indentation testing is clear-cut. As illustrated in Fig. 20.14,
during indentation, the force P and the penetration depth h are recorded as functions of
time, and the load versus displacement relationship is obtained, as shown in Fig. 20.15.
The deformation pattern of an elastic-plastic sample indentation exhibits the
following parameters:
h
f
residual hardness impression (after elastic recovery)
h
c
contact depth (i.e., depth the indenter is actually in contact with material)
h
max
calculated (projected) indentation at the peak of the load
h
s
depression of the sample around the indentation (i.e., h
s
= h h
c
)
h
p
slope extrapolated indentation depth reminder
(20.2)
FIGURE 20.13 Theoretical examples of indentations and their resulting curves. (From Bhus-
han, B. Ed., CRC Handbook of Micro/Nanotribology, 2nd ed., CRC Press, Boca Raton, FL,
1998. With Permission.)
Stress-Strain
Load-
Unloading
Cycles
Indentation
Profile
Elastic Perfectly Plastic Elastic-Plastic
True strain
Displacement
L
o
a
d
Real
h h
L
S
p
=
max
max
max
S
max
is the stiffness (inverse compliance) of the sample alone, which corresponds to
the slope of the unloading curve at the maximum load. This process begins by fitting
the unloading curve to the power law relation,
P B(h h
f
)
m
(20.3)
FIGURE 20.14 (a) Schematic of ideal tip indentation; (b) schematic of realistic indentation
with dull tip. (From Bhushan, B. Ed., CRC Handbook of Micro/Nanotribology, 2nd ed.,
CRC Press, Boca Raton, FL, 1998. With Permission.)
FIGURE 20.15 Schematic of nanoindentation load displacement behavior.
P
a
h
f
h
s
h
c
h
c
h
eff
h
Indenter shape
Surface profile
after load removal
Initial surface
Surface profile
under load
(a)
(b)
Loading
Unloading
h
max
h
f
L
o
a
d
,

P
Displacement, h
P
max
S
where P is the indentation load, h is the displacement, B and m are empirically
determined fitting parameters, and h
f
is the last displacement after complete unload-
ing, for a Berkovich indenter h
c
> h
p
. When Equation 20.3 is differentiated at the
point of maximum indentation, h h
max
, the stiffness is derived.
(20.4)
The contact depth, h
c
, is found by an approximation of the load displacement data
such that
(20.5)
In this case, P
max
is the peak indentation load and H is a geometry-dependent
coefficient so that H~for a Berkovich indenter, as shown in Fig. 20.16.
47
Other
typical tests may involve a spherical, conical, or pyramidal indenter, or some variety
of standard geometry that is impressed into the sample with the area A of the
generated deformation used to characterize the hardness.
The ratio between a corresponding load, P, and the characteristic indentation
area defines the hardness number in GPa (or kg/mm
2
) such that
H = P/A (20.6)
The force that is utilized is calculated with certainty from the machine but the
selection of an area to use for A is more difficult. The nanohardness tests can
determine behavior on a nanoscale but the indentation area is not easily visible on
the nanoscale without the aid of an SEM or an AFM.
51
A Berkovich tip is a three-sided pyramid. Three sides are chosen because the
tip mathematically intercepts at a single point. A sharply pointed tip ensures local-
ized indentation, assuming no blunting has occurred. The broad opening angle is
FIGURE 20.16 Sharp Berkovich indenter.
S
dP
dH
h h mB h h
f
m
= = =

( ) ( )
max max
1
h h
P
S
c
=
max
max
Slide
l
a
h
65.3
<0.1 m radius
Normal
Triangular base
76.9
advantageous to circumvent cutting of the surface. The contact area for a Berkovich
tip is approximated as A 0.433a
2
, where
c
a is the tip triangular base side,
and h is the tip length. This contact area is related to the actual indentation only if
the material is fully plastic. The relationship between the area and the system
compliance is called the area function. The presence of tip imperfections requires
introduction of careful but uncertain calibrations that usually provide a function of
the following form:
(20.7)
The area function is usually calibrated from indentations on hard and plastic
material surfaces, such as fused silica, in order to reduce the elastic and viscoelastic
effects of the response. However, minor imperfections due to the machine compli-
ance may be subtly included. The first term in Equation 20.7 describes a perfect
Berkovich indenter and the other terms describe deviations from Berkovich geometry
due to blunting at the tip.
52
Thurn and Cook have developed a simpler, two-parameter
form of the area function:
53
(20.8)
Their two-parameter area function acquired from the harmonic average of a
spherical tip profile and a perfect conical profile describes the Berkovich indenter
tip relatively well. They examined the function over the entire load range. The
calculated projected contact areas were verified by SEM observation.
53
The total compliance (compliance = deformation per unit force) during an
indentation experiment is related to a two-spring system in series such that C
i
and
C
s
are the compliances of the indenter and the sample and total compliance is given
by C = C
i
+ C
s
. The compliance for a Vickers and Berkovich indenter is
(20.9)
derived by King (1987), where S is the slope of the unloading curve at maximum
load and the residual elastic modulus, E
r
(assumed to be a constant), is defined as
(20.10)
In the case of indentation with an infinitely sharp, ideal Berkovich or Vickers
indenter, a relationship between the load, P, and the displacement, h, during the
24 5
2
. h
c
A h h C h C h C h C h C h
c C c c c c c
( ) . = + + + + + + 24 5
2
1
1
2
1
2
3
1
4
4
1
8
8
1
128
A
h
C h
p
c
i
i
c
i
= +
=
2
2
0
2
cot
C
S
dh
dL E A
s
r
= =

1 1
2
1
2
max
E
E E
r
s
s
i
i
=

+
1 1
2 2

loading part can also be derived, yielding
(20.11)
In general, indentations with contact depths of less than 1020% of the lm
thickness are desired in order to attain the intrinsic lm properties and to circum-
vent the substrate effect. This becomes complicated as time progresses and the
theory evolves, because the thickness of lms continues to decrease. For example,
the thickness of barrier lms currently used in semiconductor devices is well below
50 nm, and the lm stack thickness used for magnetic data storage is less than
100 nm.
54
20.3.2.4 Substrate Effect
Experimental equipment limitations pose problems in indentation. Equipment may
not possess adequate sensitivity to test at shallow indentation depths. It is quite
likely, in these circumstances, for the substrate to affect the nal measurements.
Chen and Vlassack have developed a substrate effect factor, which is essentially a
correction factor that allows one to determine the yield stress of a lm from hardness
measurements.
34
Bull has also developed a method to modify values due to the
substrate effect.
4
For a soft lm on a hard substrate (
f
/
s
1), the inuence of the
substrate is not appreciable until > h/2, being indentation depth. The yield stress
of a lm,
f
, can be measured directly from the relation H = P/A = c
b
f
, where the
constant c
b
is a constraint factor that depends on the indenter shape and material
properties. This equation is valid as long as the indentation depth is less than 50%
of the lm thickness.
During nanoindentation measurements of coated systems, the inuence of the
substrate on hardness cannot be neglected. There are several models that have been
developed to address the substrate effect.
5559
For hard thin lms, the substrate
involvement is more important than other effects. In this case, hardness of the lm
alone can be calculated from measured composite values. Among several analytical
models reported for this purpose, two were found to t experimental data well. The
Bhattacharya and Nix model
57
considers elastic-plastic deformation of a coating-
substrate system and its contribution as proportional to the volume fraction involved
in the indented area:
(20.12)
where E
f
and E
s
are the Youngs moduli, the yield strength, H
f
and H
S
the hardness
of the lm and substrate, respectively, and H
c
the composite measured hardness.
The TuckKorsunsky model is based on a dimensional analysis and a separation
of the work of indentation between the substrate and the coating. The model uses a
P
E h
E
H
H
E
r
r
r
_
,
1
]
1
2
2
1
24 5 4 .

H H H H H H E E
C S f S f S S f S f
+ ( ) exp[ ( )/( )( / ) (
/

1 2
tt t
f
/ )]
work-of-indentation approach in terms of load work partition between the coating
and the substrate and proposes the phenomenological expression for H
c
(20.13)
The model has been further modified to the form
(20.14)
where H
c
is the composite hardness, H
s
and H
f
are the substrate and film hardness,
E is the relative indentation depth, which is equal to the ratio of indentation
depth over the film thickness, N is a dimensionless transition parameter, x is a
power exponent that describes how steeply the transition occurs, and E
0
is the
value of the relative indentation depth at which the fractional hardness improve-
ment is equal to exactly 50%. Durand-Drouhin et al. used this model in a study
of TiAl(Si)N single- and multilayer thin films.
13
They concluded that crack
morphology depends on indentation depth; lateral microcracks appear at shallow
depths, while radial corner cracks occur at deeper indentation depths. Further
results by Karimi showed that brittle films have lower values of x (typically x <
3.0), while crack-resistant films exhibit higher values varying between 5 and 7.
14
In fact, x describes how steeply the transition is between the two hardnesses, H
f
and H
s
.
Plastic pile-up in the film is improved by the presence of the hard substrate.
34
However, the presence of the substrate in a hard filmsoft substrate composite
enhances the sink-in outcome. This is because the plastic region in the soft
substrate is much larger than that in the hard film and the substrate is unable to
support the large indentation load. These pile-up and substrate considerations will
be discussed below. Experimental comparisons of all of the substrate methods are
beyond the scope of this work. However, the substrate is known to play a significant
role in thin-film hardness, fracture behavior, and the degree of indentation pile-
up. Greater consideration is now given to the matter of indentation pile-up. There
are many considerations that must be reviewed when pile-up is a significant factor.
20.3.2.5 Indentation Pile-Up
Chen and Vlassak proposed a new technique to determine the contact area in a
nanoindentation experiment.
34
Their finite element results suggest that the indenta-
tion modulus of a thin film on a substrate was a function of the contact area between
the indenter and the film. If the Youngs moduli of both film and substrate are known,
then the contact area of an indentation experiment can be found by solving the
subsequent equation for the contact area A:
H H
H H
c s
f s
= +

+ 1
2
H H
H H
c s
f s
x
= +

+
1
0

(20.15)
where E is a constant and depends on the geometry of the tip. The values of E for
different indenter geometries are known. If the following equation is subsequently
used
(20.16)
and the contact area is eliminated from Equation 20.15 and Equation 20.16,
(20.17)
Saha and Nix discussed this problem and stated that the true contact pressure or
hardness of the film can be determined by accounting for the pile-up effects on the
contact area. This can be accomplished by adopting the method of Joslin and Oliver,
i.e., using the P/S
2
parameter. This parameter can be used only when the material
is elastically homogeneous and the indentation modulus is known.
35
The advantage of the P/S
2
method is that, unlike the OliverPharr technique, it
is insensitive to plastic pile-up. The parameter involves only load and stiffness values
that are directly measured during an indentation test. This is a great advantage
because it cuts out the need to measure tip shape calibration and other ambiguous
assumptions concerning material behavior.
It should be noted that this method requires knowledge of Youngs modulus of
the film. Usually the stiffness of a film can be equivalent to that of the bulk material
when the film is not very porous and the film material can be made into bulk target
material.
34
H and E are generally stable for homogeneous materials as the indentation
depth changes and this results in a constant P/S
2
with indentation depth. When P/S
2
is plotted as a function of indentation depth for fused silica, the result is a constant
curve. This result changes when a film and substrate are present, because P/S
2
will
vary with depth depending on the hardness and elastic modulus of the film and
substrate. Analyzing P/S
2
provides a more complete picture of the nanoindentation
properties of a film/substrate system than can be obtained using the OliverPharr
method alone.
The nature of the resulting indentation provides clues to the deformation prop-
erties of the film or surface, as is shown in Fig. 20.14. The indents must be analyzed
with an AFM to assess trends. Pile-up after indentation is a factor that lends
uncertainty to the aforementioned nanohardness calculations. During indentation
of a rigid plastic solid, the displaced material appears in the pile-up rim around the
periphery of the indentation site, but with elastic-plastic materials, most, if not all,
of the displaced material is accommodated by radial expansion of the elastic
S E A
R
=
2
H
P
A
P H A = =
P
S
H
E
r
2 2 2
1
4
=


surroundings, with an imperceptible change in the surface dimensions of the indented
material.
The contact point (c.p.) is assumed to be the maximum point of the pile-up, as shown
by the dotted line in Fig. 20.17. The area is then calculated by assuming that the load
supporting area of the pile-up forms an arc of radius R at the indent edges. By measuring
the indent edge a and the radius R, it is possible to calculate the total pile-up area:
(20.18)
The triangular indent area is then calculated from
(20.19)
The total projected contact area is now calculated and used in
(20.20)
to determine the film hardness.
FIGURE 20.17 (a) Projected area determination and (b) cross-section showing contact point
determination (Saha
35
).
a
R
R
c.p.
c.p.
(a)
(b)
h
p
h
o
A
R a
R
a
pile up
=
3
6 2 4
2
2
2
A a
triangle
= 0 433
2
.
H
P
A f h
c
c c
=
max
( )
The enhancement of pile-up at larger indentation depths for soft films on hard
substrates has been reported.
54
It has been suggested that the plastic flow in the film
is restricted by the relatively nondeformable substrate in a manner that causes soft
film material to flow preferentially toward the surface. This enhances the pile-up of
the material. The extent of pile-up and the depth of penetration of the indenter have
been found to depend on sample hardness. Less pile-up has been observed for films
with lower hardness and vice versa.
60
The pile-up of the material is associated with
the deformation zone. The deformation zone in harder films is smaller than that in
softer films. This may be due to the fact that harder film layers are relatively intact
and dislocation theory suggests that there are no large slip bands.
Korsunskys new approach to understanding the hardness of coated systems
helps address the pile-up issue.
59
Composite hardness is known to vary depending
on the applied load or indentation depth. Composite hardness is considered a simple
function of the relative indentation depth E (which is the indentation depth normal-
ized with respect to the coating thickness) and the substrate and coating hardnesses
in the Korsunsky model.
Figure 20.18 outlines the plastic zone that develops when the relative indentation
depth is increased and represents the primary deformation mechanisms in the inden-
tation response of a coating-substrate system dominated by coating fracture.
59
In
Fig. 20.18, Region 0 displays only an elastic response. Region I demonstrates the initial
highly localized coating elastoplasticity and fully elastic substrate before the generation
of an elastoplastic enclave in the substrate that leads to coating conformal stretching
or deformation. Radial fracture is possible along the indentation diagonals. Region IIa
is characterized by substrate deformation and conformal coating deformation sufficient
FIGURE 20.18 Representation of the primary deformation during indentation (Korsunsky
59
).
Ia
IIa
Ib
IIb
III
H
s
H
f
1 0.1 10
H
c

s
y
s
t
e
m

h
a
r
d
n
e
s
s
Relative indentation depth
0
for circumferential through-thickness fracture that grows to completion, creating an
island of dead coating. Region IIb represents the repeated circumferential or picture-
frame fracture that occurs in the coating as it is bent to conform to the plastically
deforming substrate. True substrate-only behavior occurs in Region III, where the
energy absorbed by coating, stretching, flexure, and fracture is insignificant com-
pared to the amount dissipated by substrate deformation.
59
The function developed by Korsunsky and Tuck contains a single fitting param-
eter N that describes a wide range of composite and indenter properties such as
coating plasticity, interfacial strength, and indenter geometry.
59
Additional work on pile-up is displayed in Fig. 20.19. The results were obtained
by a 2-D finite element model, which demonstrated the nature of the pile-up for
increasing values of indent load.
61
Much more work is required to quantify the
relationship between material properties with the pile-up behavior.
20.3.2.6 Continuous Stiffness Measurement Method
Another technique, continuous stiffness measurement (CSM), offers a significant
improvement in nanoindentation testing. CSM is accomplished by imposing a small,
sinusoidally varying signal on top of a DC signal that drives the motion of the indenter.
By analyzing the response of the system with a frequency-specific amplifier, data are
obtained. This allows the measurement of contact stiffness at any point along the
loading curve and not just at the point of unloading, as in conventional measurement.
The CSM technique makes the continuous measurement of mechanical properties of
materials possible in one sample experiment without the need for discrete unloading
cycles, and with a time constant that is at least three orders of magnitude smaller than
the time constant of the more conventional method of determining stiffness from the
FIGURE 20.19 Cross-sectional view of elastoplastic indentation after unloading (Raymond-
Angllis
61
).
1
5 4 3 2 1 0 1 2 3 4 5
0.8
0.6
0.4
0.2
0
0.2
X = 1
X = 1
X = 1
X = 5
X = 5
X = 10
X = 10
X = 20
X = 30
X = 60
X = 80
X = 100
X = 200
X = 1000
X = 1000
After unloading Under load
Indentation radius (mm)
I
n
d
e
n
t
a
t
i
o
n

d
e
p
t
h

(
m
m
)
slope of an unloading curve. The measurements can be made at exceedingly small
penetration depths. Thus this technique is ideal for mechanical property measurements
of nanometer-thick films. Furthermore its small time constant makes it especially
useful for measuring the properties of polymeric materials. In nonuniform materials,
such as graded materials and multilayers, the microstructure and mechanical properties
change with indentation depth. Continuous measurements of the mechanical properties
of these materials during indentation are greatly needed. Utilizing the CSM technique,
creep measurements on the nanoscale can be performed by monitoring changes in
displacement and stress relaxation. Because CSM is performed at frequencies greater
than 40 Hz, it is less sensitive to thermal drift. Load cycles of a sinusoidal shape at
high frequencies allow the performance of fatigue tests at the nanoscale.
20.3.3 ATOMIC FORCE MICROSCOPE (AFM)
An atomic force microscope (AFM) senses interatomic forces that occur between a
probe tip and a substrate. The AFM probe tip, which is similar in purpose to the
stylus on a phonograph, is normally integrated into a microfabricated, thin-film
cantilever. Once contact between the probe tip and the sample surface has been
established, the sample is translated laterally relative to the probe tip, while the
vertical position of the cantilever is monitored. Variations in sample height cause
the cantilever to deflect up or down, which changes the position sensor output. This
generates the error signal that the feedback circuit uses to maintain a constant
cantilever deflection (constant force). Normal imaging forces are in the 1- to 50-nN
range, and cantilever deflections of less than 0.1 nm can be detected.
Contact is normally considered a repulsive interaction between two surfaces,
where there is pushing on the surface followed by pushing back. This occurs only
when the distance between the surfaces is very small. Immediately outside the range
at which repulsive interactions dominate, the force between the two surfaces is
actually attractive, due to van der Waals interactions. The AFM can sense both
types of force. Repulsive forces are considered positive, and attractive forces are
considered negative. These interactions are often described by use of the Len-
nardJones potential, which gives the potential energy of two atoms separated by a
distance r.
A laser beam is trained on the back surface of the cantilever, and the reflected
beam is sent to a photodiode that is divided into two sections, A and B, as shown
in Fig. 20.20. Due to the macroscopic length of the reflected light path, any deflection
of the cantilever causes a magnified lateral displacement of the reflected laser spot
on the photodiode. The relative amplitudes of the signals from the two segments of
the photodiode change in response to the motion of the spot-the-difference signal.
The photodiode is very sensitive to cantilever deflection; detection of deflections of
<0.1 nm is readily achieved. Optical lever detection is currently used in all but a
few commercial AFM instruments. Compared to the scanning electron microscope,
AFM provides extraordinary topographic contrast in direct height measurements and
clear views of surface features with no coating necessary.
A new era in imaging began when scientists introduced a system for use of the
noncontact mode. This mode is used in circumstances where tip contact would
modify the sample in slight ways. In this mode, the tip hovers 50 to 150 above
the sample surface. Attractive van der Waals forces between the tip and the sample
are detected, and topographic images are created by scanning the tip above the surface.
Unfortunately the attractive forces from the sample are considerably weaker than the
forces used in the contact mode. Therefore the tip must be slightly oscillated so that
AC detection methods can be used to sense the small forces between the tip and the
sample by measuring the variation in amplitude, phase or frequency of the oscillating
cantilever in response to force gradients from the sample. For highest resolution, it
is necessary to measure force gradients from van der Waals forces that may extend
only a nanometer from the sample surface. Examples of AFM screens are displayed
in Fig. 20.21 and Fig. 20.22. Figure 20.22 is a typical indent trace that would be
taken to image analysis to measure the depth of the indent and the height of the pile-
up, if any occurs. This figure displays two different screens: one demonstrating pile-
up with the appropriate measurements and the other demonstrating no pile-up.
20.3.4 X-RAY DIFFRACTION (XRD)
X-ray diffraction (XRD) was performed on the PLD-deposited thin films to determine
their crystallinity and crystal structure orientation. A sample from several of the
deposition processes was investigated. A Rigaku XRD double crystal diffractometer
with Cu KD
1
radiation and a symmetrical Si (111) monochromator were used, oper-
ating in the 24 scan mode. The incident radiation source (Cu KD
1
= 0.154 nm) was
operated at 30 kV and 20 mA. The Cu KE reflection of sapphire (0006) and Si (111)
was used for calibrating the lattice constant for each substrate.
20.3.5 SCANNING ELECTRON MICROSCOPE (SEM)
The scanning electron microscope (SEM) was first described and developed in
1942.
62
Electron microscopy benefits from the wave nature of quickly moving
FIGURE 20.20 Schematic of an atomic force microscope.
Plezoelectric
sample scanner
B
A
Split (2-segment)
photo diode
Mirror
Laser
diode
Cantilever
electrons. Visible light has wavelengths from 4000 to 7000 ; however, electrons
accelerated to 10,000 KeV have a wavelength of 0.12 . This results in the greater
capabilities of the SEM as compared to conventional microscopy. Currently research
improvements in SEMs are reducing wavelength capabilities even more and thereby
increasing magnification possibilities.
The scanning electron microscope produces a beam of electrons in a vacuum that
is collimated by electromagnetic condenser lenses, focused by an objective lens, and
scanned transversely on the surface of the sample by electromagnetic deflection coils.
The chief imaging technique is accomplished by collecting the secondary electrons
that the sample releases. The secondary electrons are monitored by a scintillation
material that generates flashes of light from the electrons that are detected and
enlarged by a photomultiplier tube. The sample position is then used along with the
resulting signal to form an image that is analogous to that which would be seen
through an optical microscope. The shadowed image displays a surface topography
with a natural appearance. A schematic of the workings of the SEM is shown in
Fig. 20.23 and the SEM used for these experiments is depicted in Fig. 20.24.
20.3.6 TRANSMISSION ELECTRON MICROSCOPE (TEM)
A transmission electron microscope examines a structure by passing electrons
through a specimen. An image is formed as a shadow of the specimen on a phos-
phorescent screen. In order for electrons to pass through the specimen, it must be
very thin (usually less than 100 nm or approx. 1/25,000 in.). This requires consid-
erable preparation, equipment, time, and skill. TEM is typically used to examine
the internal structures of objects as varied as biological specimens, rocks, metals,
and ceramics.
The transmitted beam is projected onto the viewing screen, forming an enlarged
image of the slide. The image strikes the phosphor image screen and light is generated,
allowing the user to see the image. The darker areas of the image represent those
FIGURE 20.21 AFM scanned indent.
areas of the sample that fewer electrons are transmitted through (they are thicker or
denser). The lighter areas of the image represent those areas of the sample that more
electrons are transmitted through (they are thinner or less dense). The JEOL 2010
displayed in Fig. 20.25 is a high-resolution TEM with a spatial resolution of 0.194
nm. For the TEM investigations, the specimens were prepared using the conventional
sandwich technique.
FIGURE 20.22 (a) Screen image from an indent displaying no significant pile-up; (b) screen
image from an indent displaying significant pile-up.
(a)
20.4 PERFORMANCE EVALUATION OF THIN FILMS
20.4.1 APPROACH
The research included depositions characterization, mechanical testing of the films,
and finally analysis of the results. The procedure was completed in the following
steps:
1. AlN monolayer film deposition and optimization
2. TiN monolayer film deposition
3. AlN/TiN heterostructure film deposition
4. Film characterization
FIGURE 20.23 Schematic diagram of an SEM.
FIGURE 20.24 JEOL S-5000 SEM system.
Electron gun
Objective lens
Electron detector
Secondary electrons
CRT
< To vacuum
Electron beam
Stage
Vacuum chamber
5. Hardness and modulus property measurements and finite element work
6. XRD and TEM analysis
All depositions were performed on silicon (100) substrates. Before a silicon substrate
was mounted to the heater, each silicon piece was ultrasonically cleaned in acetone
for 15 min and then given an acetone vapor bath. Finally, immediately prior to the
mounting of the samples onto the heater with a thin layer of silver conductive paste,
the silicon pieces were soaked for 10 min in a weak HF etch to remove the SiO
2
layer. The Lambda Physik LPX300 PLD system used to create all the films is
displayed in Fig. 20.26.
The constant laser parameters used for this experimentation are noted in Table
20.4. These parameters resulted in a thin-film growth rate of approximately 1 /s.
The targets were also cleaned and polished before each deposition to ensure that the
track created by the laser did not interfere with the plume and subsequently the
deposition rate. Results of the target analysis are provided in Section 20.5. The pulse
repetition rate was set at the standard 10 Hz.
20.4.2 AlN MONOLAYER FILM DEPOSITION AND OPTIMIZATION
The laser energy was varied from 300 mJ to 750 mJ. When the gas for the laser was
newly filled, it had the capacity to reach higher power values, but over time the laser
energy capacity diminished. The first tests changed the laser energy, while holding
constant the substrate temperature, the ambient gas pressure, and the target material.
The aperture through which laser beam passed was 12 mm u 6 mm. The AlN target
was mounted, and the vacuum was started. After a vacuum of 5 to 8 u10
6
torr was
achieved, the laser was turned on. After a warm-up period, pre-ablation of 500 pulses
was run. The shield over the sample was then removed and 20,000 pulses at 10 Hz
were performed. This test was similarly repeated at 300, 400, 500, 600, 650, and
FIGURE 20.25 JEOL 2010-F TEM.
700 mJ. The thickness of the resultant film was measured by the Wyko profileometer
and the results are discussed in Section 20.5.
This process was repeated in a similar fashion with the laser energy constant
and the ambient gas pressure and the substrate temperature varied. All the results
are presented in Section 20.5.
20.4.3 TiN MONOLAYER FILM DEPOSITION
The TiN monolayer films were deposited at optimal process conditions with changes
in only the substrate temperature. Tests were performed in the fashion previously
described: 20,000 pulses at 10 Hz were used to deposit the films at 200C, 500C,
600qC, 700C, and 800C. The results of the tests are covered in Section 20.5. A
value of 20,000 pulses was selected as the standard total.
20.4.4 AlN/TiN HETEROSTRUCTURE DEPOSITION
The depositions of AlN/TiN were used to create multilayers whose periods varied
from 2.4 nm to 140 nm. A constant film thickness of ~250 nm for all depositions
FIGURE 20.26 Lambda Physik LPX300 PLD system.
TABLE 20.4
Constant Excimer Laser Parameters Used
for Thin-Film PLD of All Films Deposited
Parameter Value
Wavelength 248 nm
Pulse repetition rate 10 pulses per second
Pulse duration 30 nsec
was the goal. The samples created are discussed in Section 20.5. The total number
of pulses was held constant at 20,000. The hardness and modulus were found via
the CSM method on an MTS nanoindenter as discussed in Section 20.3.
20.4.4.1 Hardness and Modulus versus
Deposition Temperature
The temperature was varied from 200C to 800C for the different multilayered
films. The sample deposition parameters and properties are explained in Section
20.5. An annealing study was performed on a multilayered, low-temperature
(200C) deposited sample. The sample was hardness tested. It was then annealed
at 500C for 2 h and the hardness was retested. The sample was then annealed at
800C for 2 h and the hardness was tested again. The hardness plots are displayed
in Section 20.5.
20.4.4.2 Hardness versus Depth of Indentation
Hardness tests were performed at various different indentation loads, therefore
creating different indentation depths. The purpose of this study was to use the Nix
and Korsunsky models and determine their appropriateness for the AlN/TiN material
system and the substrate effect. A TiN film was analyzed in greater detail to discern
the correct model for describing the behavior. Further studies were performed to
determine the constants needed for the NixBhattacharya and TuckKorsunsky
models discussed in Section 10.3. These models were then compared for their
compatibility with the current data. The results are discussed in Section 10.5.
The hardnesses of the various multilayered samples were compared in relation to
their period thickness O7he results are presented in Section 20.5.
AFM imaging of the indents was performed to determine the extent and nature of
the pile-up found around the indents. The line profiles of these images were deter-
mined and analyzed to discover the nature of the pile-up and assess the need for
explanation of the pile-up and the substrate effect. Images and line scans are dis-
played in Section 20.5. The material parameters were used in an LS DANA finite-
element program to model the indent plastic behavior. The results are presented in
Section 20.5.
20.4.5 X-RAY DIFFRACTION
X-ray diffraction (XRD) was performed on the pulsed laser-deposited thin films to
determine their crystallinity and crystal structure orientation. A few samples were
tested to ascertain the quality of the deposition and the effect the deposition tem-
perature had on the crystallinity. A Rigaku XRD double-crystal diffractometer with
Cu KD
1
radiation and a symmetrical Si (111) monochromator was used, operating
in the 24 scan mode. The incident radiation source (Cu KD
1
0.154 nm) was
operated at 30 kV and 20 mA. The Cu KE reflection of silicon (100) was used for
calibrating the lattice constant for the substrate. Results are displayed in Section 20.5.
20.4.6 TRANSMISSION ELECTRON MICROSCOPY (TEM)
Samples were prepared according to the detailed specifications in Chapter 3. Sample
B33, which was an AlN/TiN ten-layer film deposited at 500C, was analyzed in the
JEOL 2010-F TEM under different conditions. Results are presented in Section 20.5.
20.5 OPTIMIZATION OF RESULTS
20.5.1 LASER ENERGY EFFECT RESULTS
Laser energy studies were performed and the results are shown in Fig. 20.27. The
temperature (500C) and the ambient gas pressure (15 mtorr) were held constant
for each laser energy level. At each deposition condition, the quality of the resultant
film was examined and the film thickness was quantified as a means of quality
analysis. Notice the peak in film thickness at an energy range of approximately 600
mJ. This is the energy chosen for the remainder of the tests.
20.5.1.1 Ambient Gas Pressure Effects
When nitrides are deposited, the process should occur in a nitrogen-rich environment.
However, as shown in Fig. 20.5, there is an interaction of the plume with the ambient
gas. These collisions with the inert gas change the kinetic energy distribution of the
deposited particles. The data from this work verify the models previously introduced.
The ambient gas pressures used were 0, 15, 50, 135, and 255 mtorr of N
2
. Laser
energy was constant at 400 mJ, temperature was constant at 500C, and the number
of pulses was also constant at 20,000. As shown in Fig. 20.28, hardness tends to
decrease as the internal gas pressure and, in turn, the number of plumegas collisions
increase. Samples had a constant temperature, 500qC, constant laser energy 400 mJ,
and constant number of pulses. The chosen ambient gas pressure was 15 mtorr,
which diminished the effect of the gas collisions.
20.5.1.2 Location of Deposition on Heater
The size and shape of the plume have been discussed. A study of the thickness along
the entire width of the heater was performed. The results of step thickness measure-
ments across the 2-in. width of the heater are displayed in Fig. 20.29. Each data
point is an average of four step measurements. The thickness steps were measured
across the sample shown in the smaller box with the start and end locations shown
by the arrows. The results verify the work of other researchers who have achieved
a 1-cm optimum deposition area.
20.5.2 AlN MONOLAYER DEPOSITION AND PROPERTIES
AlN films were deposited at 200C, 500C, and 800C for this investigation. Initial
hardness tests were run on the monolayer films of the AlN. Hardness and modulus
of samples were determined by means of depth-sensing nanoindentation using the
continuous stiffness measurement methods. An example of an indent from the AlN
FIGURE 20.27 Graph of thickness of the AlN film versus laser energy.
FIGURE 20.28 Nanohardness of AlN monolayer films at different nitrogen pressures.
T
h
i
c
k
n
e
s
s

o
f

f
i
l
m

(
n
m
)
Film thickness vs. laser energy
Energy of laser (mJ)
300 400 500 600 700
340
320
300
280
260
240
220
200
180
160
140
120
100 200 300 400 500 0
0
2
4
6
8
10
12
14
16
18
Nanohardness tests
samples with varying ambient N
2
pressure
N
a
n
o
h
a
r
d
n
e
s
s

(
G
P
a
)
No N
2
pure vacuum
15 m torr N
2
50 m torr N
2
135 m torr N
2
255 m torr N
2
indentation is shown in Fig. 20.30. This film was deposited at 500C with 15 mtorr
ambient N
2
and on a Si (100) substrate.
The hardness (~16 GPa) and the modulus (190 GPa) of the AlN monolayer film
were below the values published from earlier work (Rawdanowicz). The film is very
thin; therefore the 10% rule for indentation testing is violated. The substrate affects the
hardness to a great degree. Figure 20.31 displays the hardness and elastic modulus of
the monolayer AlN film. The thickness is ~250 nm and is deposited on Si (100) at 500C.
20.5.3 TiN MONOLAYER DEPOSITION AND PROPERTIES
TiN film samples were deposited at 200C, 500C, 700C, and 800C with 20,000
pulses. Initial hardness tests were run on the monolayer films of the AlN. The
hardness and modulus of the samples were determined by means of depth-sensing
nanoindentation using continuous stiffness measurement methods. An example of
the indent from the TiN indentation is shown in Fig. 20.32. This film was deposited
at 500C.
The shapes of the load displacement curves for two representative TiN samples
are displayed in Fig. 20.33. The solid line was deposited at 200C, and the dashed
line was deposited at 800C. The shape of the unloading curve was examined in
detail to verify the application of the power law relationship given in Equation 20.3.
The initial portion of the unloading curve from Fig. 20.33 is expanded and plotted
in Fig. 20.34. The unloading segment is dominated by elastic behavior. The dis-
placements were shifted laterally by subtracting the final depth, h
f
, for each data
point by the total depth, h, to force all of the curves to pass through the origin. The
fact that the data is linear on logarithmic axes implies that the unloading curves are
appropriately described by a power law relation.
FIGURE 20.29 Thickness measurements across the face of the heater.
160
140
120
100
80
60
40
20
0
T
h
i
c
k
n
e
s
s

(
m
m
)
0 5 10 15 20 25 30 35 40 45 50
Distance profile from one edge of the heater (mm)
Thickness profile of all thin film
1 cm
The jog in the unloading portion of Fig. 20.33 is evidence of a large crack
formed during unloading. When the load on the indent was removed, the stresses
were relaxed, and the crack appeared as a discontinuity in the load displacement
curve. The shallower slope of the 200C sample displayed in Fig. 20.34 verifies its
soft behavior. The resulting variation of the hardness and elastic modulus as a
function of indentation depth is displayed in Fig. 20.35. This film was deposited
at 500C and indented with the Berkovich pyramid indenter. The primary influence
of the Berkovich indenter tip with radius of approximately 60 nm is the rapid
increase in the hardnesses and elastic moduli that occurs in the first 40 nm to
60 nm of the indentation. The indenter, in most cases, broke through the films and
contacted the substrate. The thickness of the thin films located directly under the
plume was greater than that in the area located at the plumes edge. Therefore exact
thicknesses of the pulsed laser-deposited thin films were unknown. The films were
too thin to restrict the indentation depth to approximately 10% of the film thickness.
The TiN displayed hardness comparable to the hardnesses in Table 20.3, but the
modulus was below what was expected. The substrate effect is the likely cause of
the lower modulus because the film is thin and the plastic zone from the indenter
would almost immediately extend into the substrate. This affects the elastic modulus
to a degree.
20.5.4 AlN/TiN MULTILAYERED FILM PROPERTIES
Nanoindentation tests were performed on the multilayered films using an MTS
Nanoindenter
XP by the process explained in previous chapters.

FIGURE 20.30 SEM micrograph of an indent on AlN monolayer film.
20.5.4.1 Hardness and Modulus versus Deposition Temperature
As the temperature of the deposition increased, there was a difference in the film
characteristics. This difference was first noted in the values of nanohardness and
subsequently in the SEM images and cracking pattern of indents studied during
FIGURE 20.31 (a) Nanohardness and (b) elastic modulus results of monolayer AlN film.
0 50 100 150 200 250 300 350
6
8
10
12
14
16
18
N
a
n
o
h
a
r
d
n
e
s
s

(
G
P
a
)
AlN monolayer film
(a)
0 50 100 150 200 250 300 350
50
100
150
200
AlN monolayer film
E
l
a
s
t
i
c

m
o
d
u
l
u
s

(
G
P
a
)
(b)
fracture mechanics analysis. The temperature of the deposition is known to affect
the atomic behavior of the film (i.e., whether the films are crystalline or amorphous).
If the film is crystalline, then the grain size is also affected by the deposition
temperature. The higher the temperature of the deposition, the larger the grains will
be, because at higher temperatures the atoms have greater energy to move and
rearrange once on the target. As seen in Fig. 20.36 through Fig. 20.39, the highest
FIGURE 20.32 SEM micrograph of an indent on a TiN film.
FIGURE 20.33 Load versus indentation depth plots for TiN thin films deposited at two
different temperatures.
0 100 200 300 400 500 600
0
10
20
30
40
50
60
L
o
a
d

(
m
N
)
200C
800C
Load vs indenter displacement
for two different deposition
conditions of TiN thin film
Indentation depth (nm)
deposition temperature does not always produce the hardest film. The highly energetic
nature of the laser deposition also leads to unique properties. The paramount tem-
perature for optimal-quality film growth occurs when there is adequate surface
diffusion to permit surface atoms to minimize their surface energy, and therefore
become thermodynamically stable.
43
The current tests demonstrate that, for tests
containing TiN as the outer layer, a deposition temperature between 700C and
800C produces the hardest film. Further XRD analysis, depicted in Fig. 20.45
through Fig. 20.48, shows that the higher temperature film was indeed of a crystalline
nature and that the low temperature deposit was amorphous.
All of the hardness plots demonstrated a convergence to the Si (100) substrate
hardness (12.75 GPa) or modulus (180 GPa). The increase in properties was not a
linear increase but all curves displayed a trend toward greater hardness and modulus
as the deposition temperature increased.
The hardnesses of various samples were compared to their respective period thick-
nesses O and plotted in Fig. 20.40. Prior research has shown a trend toward a harder
film as the period decreases.
4
As stated in Chapter 1, the properties of multilayer
coatings vary as a function of O. Bull found the hardness of the structural multilayers
was less than those of comparable single-layer films at low O However, it increased
up to a maximum at O~10 nm. The hardness of very thick structural multilayers was
indistinguishable from the average value of single-layer coating hardness.
4
Prior
research has shown that the TiN layers in the TiN/AlN superlattice manipulate the
structures of the AlN layers when TiN/AlN films are created as a superlattice. At a
superlattice period of 2.5 nm, the AlN layers take on the same structure as the TiN
(i.e., the NaCl structure
7,8
). AlN forms in a hexagonal structure during equilibrium,
FIGURE 20.34 Data from Figure 20.33, adjusted for unloading, plotted on a log scale.
10
Unloading curves
L
o
a
d

(
m
N
)
Displacement, h-h
f
(nm)
100
800C
200C
but the TiN acts in this circumstance as a model and forces AlN into the NaCl
structure.
The graph in Fig. 20.40 was created after testing many films created by the PLD
process. The data demonstrate that multilayering appears to only slightly affect
hardness, while it significantly delays the occurrence of indentation-induced large
cracks. This can be deduced from the higher values of the fracture toughness. The
FIGURE 20.35 (a) Nanohardness and (b) elastic modulus plots for TiN monolayer film.
0 100 200 300 400 500
14
16
18
20
22
24
26
28
30
(a)
TiN monolayer film
N
a
n
o
h
a
r
d
n
e
s
s

(
G
P
a
)
0 50 100 150 200 250 300 350 400 450 500
360
340
320
300
280
260
240
220
200
180
160
(b)
TiN monolayer film
M
o
d
u
l
u
s

o
f

e
l
a
s
t
i
c
i
t
y

(
G
P
a
)
smallest O, the 2-nm curve, is not the hardest material. This result is consistent with
Bulls findings.
FIGURE 20.36 Nanohardness plots for bilayer AlN/TiN films.
FIGURE 20.37 Elastic modulus plots for bilayer AlN/TiN films.
50 100 150 200 250 300 350 400 450 500
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
AlN-TiN 2-layer film
temperature variance
N
a
n
o
h
a
r
d
n
e
s
s

(
G
P
a
)
200C
500C
400C
300C
600C
700C
800C
0 200
40
60
80
100
120
140
160
180
200
220
240
260
280
300
E
l
a
s
t
i
c

m
o
d
u
l
u
s

(
G
P
a
)
200C
400C
500C
600C
800C
AFM imaging of the indents was used to determine the extent and nature of the
pile-up around the indents. In Fig. 20.41, pile-up in the AlN/TiN film is evident,
FIGURE 20.38 Nanohardness plots for four-layered AlN/TiN films.
FIGURE 20.39 Elastic modulus plots for four-layered AlN/TiN films.
0 100 200 300 400 500 600
8
9
10
11
12
13
14
15
16
17
18
N
a
n
o
h
a
r
d
n
e
s
s

(
G
P
a
)
200C
500C
600C
700C
800C
50 100 150 200 250 300 350 400 450 500 550 600
800C
700C
600C
500C
200C
230
220
210
200
190
180
170
160
150
140
130
E
l
a
s
t
i
c

m
o
d
u
l
u
s

(
G
P
a
)
and the line scan of the indent is displayed. This was an AlN/TiN ten-layer film
deposited at 700C, with a 150-mN indent load. Figure 20.42 displays a repre-
sentative image without significant pile-up deposited at 800C, 30-mN indent
load.
The true contact pressure or hardness of a film can be determined by accounting
for the pile-up effects on the contact area. This can be accomplished by adopting
the method of Joslin and Oliver (i.e., using the P/S
2
parameter). The P/S
2
parameter
can be used only when the material is elastically homogeneous and the indentation
modulus is known. It is necessary to choose a filmsubstrate combination that is
elastically homogeneous for a film on a substrate,
(20.17)
The true contact pressure or hardness may be determined by measuring the load
and contact stiffness, even when pile-up occurs. Joslin and Oliver were the first to
use the P/S
2
parameter in a study of the mechanical properties of ion-implanted Ti
alloys. They suggested that the P/S
2
parameter is a direct measure of true hardness
since it is proportional to H. This parameter is most accurate when the film acts as
a part of a homogeneous system with the substrate. While not the most accurate,
information is also gleaned from nonhomogeneous systems. These calculations will
be applied to the current data in the future.
Figure 20.43 shows the pile-up area measured by AFM. The area of pile-up was
found mathematically using an integral program from Origin Software. The area
was found to increase with an increasing penetration load as similarly found by
FIGURE 20.40 Nanohardness of films deposited at different conditions.
200 300 400 500 600 700 800
2
4
6
8
10
12
14
16
18
20
22
24
26
N
a
n
o
h
a
r
d
n
e
s
s

(
G
P
a
)
Temperature of deposition (C)
2 nm layers
6 nm layers
24 nm layers
60 nm layers
120 nm layers
P
S
H
E
r
2 2 2
1
4
=


Chen and Beegan.
34,63
Figure 20.44 demonstrates the nature of the pile-up for
increasing values of the indent load and is a direct verification of 2-D finite element
modeling by Ramond-Angllis, as shown in Fig. 20.21.
20.5.5 X-RAY DIFFRACTION (XRD) DATA
First, comparisons were made between two TiN samples deposited at different
temperatures. In Fig. 20.45, the Bragg-Brentano 24 scans show that the TiN film
grown on silicon (100) at 800C is highly textured, as there are only (h00) peaks
FIGURE 20.41 AFM trace showing pile-up height and indent depth.
FIGURE 20.42 AFM trace showing no significant pile-up. AlN/TiN ten-layer film.
with h = 2 and 4. The smaller peak height of the Ti (400) line is in perfect
accordance with the theoretical integrity (8%) of the (400) line with respect to a
100% peak of the (200) line. Figure 20.46, which displays a film deposited at
200C, displays none of the crystalline peaks evident in Fig. 20.45. XRD data
from the two figures verifies that the structure of Fig. 20.46 is a completely
FIGURE 20.43 Pile-up area calculated from line profile of an AFM image.
FIGURE 20.44 Graph of pile-up height versus indent load.
50 100 150 200
0
50
100
150
200
250
300
350
Pile-up area calculated vs indent load
Indent load (mN)
P
i
l
e
-
u
p

a
r
e
a

(
n
m
2
)
D
Data 7D
600
2 0 2 4 6 8 10 12 14 16 18 20 22
500
400
300
200
100
0
100
200
D
e
p
t
h

o
f

I
n
d
e
n
t

(
n
m
)
200 mN
150 mN
150 mN
80 mN
40 mN
30 mN
20 mN
15 mN
10 mN
om
amorphous TiN film, whereas the film deposited at 800C is crystalline. The film
displayed in Fig. 20.47 is an AlN monolayer tested at Rigaku laboratories with
the Rigaku Ultima III equipment. The peaks are similarly located as those in Fig.
20.48, which displays a multilayered AlN/TiN film with a layer thickness, O~20
nm. The peaks at ~33 and 37 are consistent with values published for hexagonal
AlN (JCPDS).
The film density and roughness measurements from the Rigaku Ultima III XRD
machine are displayed in Table 20.5. These measurements also calculate the film
density as 3.235 g/cm
3
, which is consistent with values in the literature for a
hexagonal AlN structure specified in Chapter 1. The film was deposited at 500C,
so strong crystalline peaks were not expected and in fact are not evident.
20.5.6 TEM DATA
Figure 20.49 displays the results of the high-resolution STEM-Z contrast of the
cross-section. This image exhibits the substratefilm interface along with the layer
thickness. Notice the overall film structure and the relative layer thickness. The
darker areas of the film are the TiN layers, which are narrower than the AlN layers.
Figure 20.50 presents the cross-sectional STEM-Z micrograph at a greater mag-
nification. The crystalline nature of the substrate and the amorphous AlN first
layer is evident. The bright areas of the substrate are representative of the Si (100)
orientation.
Figure 20.51 displays the film cros-section, with the darker area inhabited by
the lighter elements. TiN, the heaviest element, emerges as the brighter layer. The
first layer is AlN, but it is darker than the intermediate AlN layers. The TiN has
FIGURE 20.45 XRD analysis of TiN monolayer film deposited at 800C on Si (100).
20 30 40 50 60 70 80 90 100
0
200
400
600
800
1000
TiN monolayer film 800C
X
-
r
a
y

i
n
t
e
n
s
i
t
y
Angle 2s
T
i
N

(
2
0
0
)
T
i
N

(
4
0
0
)
S
i

(
4
0
0
)
likely diffused into the other AlN layers and they therefore appear lighter than the
first layer. The first layer grows epitaxially on the Si substrate and is therefore
more stable.
FIGURE 20.46 XRD analysis of TiN monolayer film deposited at 200C on Si (100).
FIGURE 20.47 XRD analysis of AlN monolayer film deposited at 800C on Si (100).
20 30 40 50 60 70 80 90 100
0
200
400
600
800
1000
TiN monolayer film 200C
X
-
r
a
y

i
n
t
e
n
s
i
t
y
Angle 2s
S
i

(
4
0
0
)
20 30 40 50 60 70 80 90
0
200
400
600
800
AIN monolayer film 800C
I
n
t
e
n
s
i
t
y
/
c
o
u
n
t
s
2s degrees
A
I
N

(
1
0
0
)
A
I
N

(
0
0
2
) S
i

s
u
b
s
t
r
a
t
e
20.6 CONCLUSIONS
The PLD process was optimized for AlN/TiN thin films. The optimum laser energy
was discovered to be approximately 600 mJ, and the optimum ambient N
2
pressure
was found to be 15 mtorr. Pulsed laser-deposited films can be created in this system
with maximum thickness in an area 1 cm
2
in size. The hardness values and fracture
morphology were reported for these pulsed laser-deposited multilayer AlN/TiN films
deposited on Si (100). The temperature of the substrate during the deposit was found
to play an important role in the crystalline characteristics and mechanical properties.
The hardness and elastic modulus values for pulsed laser-deposited multilayer
AlN/TiN films deposited on Si (100) have been reported. The hardnesses for many
2.5- to 120-nm AlN/TiN bilayers deposited on silicon substrates concurred with
recently reported results. However, these multilayer hardness values of 18 to 23 GPa
were notably less than the typically expected rule-of-mixture values for AlN and
TiN thin-film component values of 24 GPa and 35 GPa, respectively. No strong
FIGURE 20.48 XRD analysis of AlN/TiN ten-layer film deposited at 500C on Si (100).
TABLE 20.5
Reflectivity Analysis Results
Layer Density (g/cm
3
) Thickness (nm) Roughness (nm)
AlN oxide 2.286 2.24 0.621
AlN 3.235 235 0.154
20 30 40 50 60 70 80 90 100
600
550
500
450
400
350
300
250
200
150
100
50
0
AIN-TiN 10-layer film 500C
X
-
r
a
y

i
n
t
e
n
s
i
t
y
Angle 2s
A
I
N

(
1
0
0
)
A
I
N

(
0
0
2
)
S
i

(
4
0
0
)
S
i

(
2
0
0
)
correlation between the individual layer thickness and the hardness was found. The
data show that multilayering appears to slightly affect hardness but significantly
delay the occurrence of indentation-induced large cracks. This can be deduced from
the higher values of the fracture toughness.
When annealing was performed on the low temperature-deposited sample, the
annealed hardness did not reach close to the hardness of the higher temperature-
deposited sample. This was analyzed to mean that film quality is not just a result
of temperature. Heat must be combined with the action of the highly energetic plume
FIGURE 20.49 High-resolution STEM-Z contrast image of AlN/TiN ten-layer film deposited
at 500C.
FIGURE 20.50 Cross-sectional TEM image of Sample B33 exhibiting crystalline Si (100)
substrate adjacent to the mainly amorphous AlN first layer.
hitting the substrate surface. In other words, shortcuts during deposition cannot be
easily addressed after deposition.
The continuous nanoindentation stiffness measurement method for measuring
the hardness and elastic modulus provides sound hardness measurement results. This
method is optimal for thin-film measurement. The substrate effect is well docu-
mented and evident in this body of work. The XRD analysis verified the crystal
natures of the hardest films that were created at the higher temperature range 700
to 800qC. In order to create the hardest films, deposition must occur in the 700 to
800qC range. The AlN XRD data verified that the AlN was of a hexagonal structure
with a density of 3.235 g/cm
3
. Initial TEM work demonstrated the exact layer
thickness, the lack of crystallinity at 500C, and the possible diffusion of TiN into
the AlN via the Z-contrast image.
PROBLEMS
1. Compare the mechanical properties of thin film and bulk of the same
material system.
2. Describe the role of growth temperatures on the mechanical and the
structural properties of AlN and TiN heterostructures.
3. Discuss the limitations of nanoindentation technique in the measurement
of hardness and elastic moduli of thin films.
FIGURE 20.51 Cross-sectional STEM-Z micrograph of AlN/TiN layers deposited on an Si
(100) substrate.
REFERENCES
1. Holleck, H. and Schulz, H., Thin Solid Films, 153, 11, 1987.
2. Holleck, H. and Schier, V., Surf. Coat. Technol., 7677, 328, 1995.
3. Holleck, H., J. Vac. Sci. Technol A, 4, 6, 2661, 1986.
4. Bull, S.J. and Jones, A.M., Surf. Coat. Technol., 78, 173184, 1996.
5. Helmersson, U., Todorova, S., Marbert, L., Barnett, S.A., Sundgren, J.E., and Geen,
J.E., J. Appl. Phys., 62, 481, 1987.
6. Nakonechna, O., Cselle, T., Morstein, M., and Karimi, A., Thin Solid Films, 2003.
7. Sproul, W.D., Surf. Coat. Technol, 81, 1, 1996.
8. Elshabini-Riad, A., Barlow, F., Eds., Thin Film Technology Handbook, McGraw Hill,
1998.
9. Sundgren, J.-E., Thin Solid. Films, 128, 21, 1985.
10. Sundgren, J.-E. and Hentzell, H.T.G., J. Vac. Sci. Technol. A, 5, 2259, 1986.
11. Sundgren, J.E., Birch, J., Hakansson, G., Hultman, L., and Helmersson, U., Thin
Solid Films, 193194, 818, 1990.
12. Veprek, S. and Reiprich, S., Thin Solid Films, 268, 64, 1995.
13. Durand-Drouhin, O., Santana, A.E., Karimi, A., Derflinger, V.H., and Schutze, A.,
Surf. Coat. Technol., 163164, 260266, 2003.
14. Karimi, A., Wang, Y., Cselle, T., and Morstein, M., Thin Solid Films, 420421, 275,
2002.
15. Koehler, J.S., Phys. Rev. B, 2, 547, 1970.
16. Yoshii, K., Takagi, H., Umeno, M., and Kawabe, H., Metall. Trans. A, 15, 1273, 1984.
17. Tsakalokos, T. and Hilliard, J.E., J. Appl. Phys., 54, 734, 1983.
18. Mirkarimi, P.B., Hultman, L., and Barnett, S.A., Appl. Phys. Lett., 57, 2654, 1990.
19. Baumvol, I.J.R., Stedile, F.C., Schreiner, W.H., Freire, Jr., F.L., and Schrer, A., Surf.
Coat. Technol., 59, 187, 1993.
20. Godbole, V.P. and Narayan, J., J. Mater. Res., 11, 1810, 1996.
21. Bathe, R., Vispute, R.D., Habersat, D., Sharma, R.P., Venkatesan, T., Scozzie, C.J.,
Ervin, M., Geil, B.R., Lelis, A.J., Dikshit, S.J., and Bhattacharya, R., Thin Solid
Films, 398399, 575, 2001.
22. Talyansky, V., Vispute, R.D., Ramesh, R., Sharma, R.P., Venkatesan, T., Li, Y.X.,
Salamanca-Riba, L.G., Jones, K.A., and Iliadis, A.A., Thin Solid Films, 323, 37, 1998.
23. Pearton, S.J., Wide Bandgap SemiconductorsGrowth, Processing and Applications,
Noyes Corporation, 2003.
24. Manaila, R., Biro, D., Barna, P.B., Adamik, M., Zavaliche, F., Craciun, S., and
Devenyi, A., Appl. Surf. Sci., 91, 295, 1995.
25. Toth, L.E., Transition Metal Carbides and Nitrides, Academic Press, New York, 1971.
26. McKenzie, D.R., Yin, Y., McFall, W.D., and Hoang, N.H., J. Phys. Condens. Matter,
8, 5883, 1996.
27. Patsalas, P. and Logothetidis, S., J. Appl. Phys., 93, 989, 2003.
28. Bae, Y.W., Lee, W.Y., and Blau, P.J., Appl. Phys. Lett., 66, 1895, 1995.
29. Bergman, E., Kaufmann, H., Schmid, R., and Vogel, J., Surf. Coat. Technol., 42,
237251, 1990.
30. Munz., M.D., J. Vac. Sci. Technol A, 4, 2717, 1986.
31. Shu, N., Kumar, A., Alam, M.R., Chan, H.L., and You, Q., Appl. Surf. Sci., 109, 366,
1997.
32. Nose, M., Nagae, T., Yokota, M., Saji, S., Zhou, M., and Nakada, M., Surf. Coat.
Technol., 116119, 296, 1999.
33. Chou, W., Yu, G., and Huang, J., Surf. Coat. Technol., 149, 7, 2002.
34. Chen, Xi and Vlassak, J.J., J. Mater. Res., 19, 2974, 2001.
35. Saha, R. and Nix, W.O., Mater. Sci. and Eng. A., 319321, 898901, 2001.
36. Auger, M.A., Sanchez, O., Ballesteros, C., Jergel, M., Aguilar-Frutis, M., and Falcony,
C., Thin Solid Films, 433, 211, 2003.
37. Singh, R.K. and Narayan, J., Phys. Rev. B, 41, 8843, 1990.
38. Zhigilei, L.V., Kodali, P.B.S., and Garrison, B.J., Chem. Phys. Lett., 276, 269, 1997.
39. Fahler, S., Stormer, M., and Krebs, H.U., Appl. Surf. Sci., 109110, 433, 1997.
40. Sharma, A.K., Thareja, R.K., Willer, U., and Schade, W., Appl. Surf. Sci., 206, 137,
2003.
41. Tabbal, M., Mrel, P., Chaker, M., El Khakani, M.A., Herbert, E.G., and Lucas, B.N.,
Surf. Coat. Technol., 116119, 452, 1999.
42. Gottmann, J. and Kreutz, E.W., Surf. Coat. Technol., 116, 1189, 1999.
43. Sturm, K., Fahler, S., and Krebs, H.U., Appl. Surf. Sci., 154155, 462466, 2000.
44. Geohegan, D.B., Thin Solid Films, 220, 12, 138, 1992.
45. Hubler, G.K., Chrisey, D.B., and Hubler, G.K., Eds., Pulsed Laser Deposition of Thin
Films, John Wiley, New York, 1994, p. 327.
46. Timm, R., Willmott, P.R., and Huber, J.R., Appl. Phys. Lett., 71, 1966, 1997.
47. Liu, X., Yin, J., Liu, Z.G., Yin, X.B., Chen, G.X., and Wang, M., Appl. Surf. Sci.,
174, 35, 2001.
48. Tabor, D., The Hardness of Metals, Oxford University Press, London, 1951.
49. Oliver, W.C. and Pharr, G.M., J. Mater. Res., 7, 1564, 1992.
50. Doerner, M.F. and Nix, W.D., J. Mater. Res., 1, 601, 1986.
51. Cheng, Y.T. and Chen, C.M., Appl. Phys. Lett., 73, 5, 614, 1998.
52. Linchinchi, M., Lenardi, C., Haupt, J., and Vitali, R., Thin Sol. Films, 312, 240, 1998.
53. Thurn, J. and Cook, R.F., J. Mater. Res., 17, 1143, 2002.
54. Tsui, T.Y., Vlassak, J., and Nix, W.D., J. Mater. Res., 14, 2196, 1999.
55. Buckle, H., Science of Hardness Testing and Its Research Applications, Westbrook,
J.H. and Conrad, H., Eds., ASM, Metals Park, OH, 1973, p. 453.
56. Fabes, B.D., Oliver, W.C., McKee, R.A., and Walker, F.J., J. Mater. Res., 7, 3056,
1992.
57. Bhattacharya, A.K. and Nix, W.D., Int. J. Sol. Struct., 24, 1287, 1998.
58. Korsunsky, A.R., McGurk, M.R., Bull, S.J., and Page, T.F., Surf. Coat. Technol., 99,
12, 171, 1998.
59. Tuck, J.R., Korsunsky, A.M., Bhat, D.G., and Bull, S.J., Surf. Coat. Technol., 139,
63, 2001.
60. Barshilia, H.C. and Rajam, K.S., Surf. Coat. and Technol., 155, 23, 195, 2002.
61. Raymond-Angllis, C., Ecole Nationale Suprieure des Mines de Paris, 1998.
62. Zworykin, V.K., Hiller, J., and Snyder, R.L., ASTM Bull., 15, 117, 1942.
63. Alexandrou, I., Scheibe, H.J., Kiely, C.J., Popworth, A.J., Amaratunga, G.A.J., and
Schultrich, B., Carbon Films with an sp
2
Network Structure, Phy. Rev. B, 60, 10903,
1999.
64. Joslin, D.L. and Oliver, W.C., A New Method for Analyzing Data from Continuous
Depth-Sensing Microindentation Tests, Journal of Materials Research, 5(1): 123126,
Jan. 1990.
21
Polarization in
Nanotubes and
Nanotubular Structures
Marco Buongiorno Nardelli,
Serge M. Nakhmanson, and Vincent Meunier
CONTENTS
21.1 Introduction
21.2 Modern Theory of Polarization
21.2.1 Computing Polarization Using Berry Phase Method
21.2.2 Polarization from Wannier Functions
21.2.3 A Simple Way to Compute Berry Phases on a Computer
21.3 Computational Details
21.4 Polarization in Nanotubes
21.4.1 Berry Phase Method
21.4.2 Maximally Localized Wannier Functions
21.5 Piezoelectricity in Nanotubes
21.6 Polarization Effects in Nanotubular Structures
21.6.1 Nanotube Superlattices and Heterostructures
21.6.2 Multiwalled Hybrids
21.6.3 Experimental Results
21.7 Conclusions and Future Perspectives
Problems
Acknowledgments
References
21.1 INTRODUCTION
A material is known as polar if, in one way or another, it displays macroscopic
electric polarization. All polar materials exhibit piezoelectricity: they develop non-
zero electric polarization when strained and change shape when placed into an
external electric field. Structures that have nonzero intrinsic polarization also
called spontaneous polarization even when no strain is applied to them are known
as pyroelectrics. The prefix pyrofrom the Greek word for firewhich is used to
indicate the permanent electric moment associated to these materials, arises from
the fact that pyroelectricity was first discovered in certain crystals when they were
warmed. However, it was soon established that these materials also had permanent
dipole moments at lower temperatures as well, only they were hidden by the
adsorption of polar ions on the faces of the samples. Once these ions were desorbed
at high temperatures, the permanent moment would appear. Usually the existence
of pyroelectricity is attributed to the reduced symmetry in the geometrical structure
of certain crystalline materials. In practice, if we model a crystal as a superposition
of microscopic dipoles, these dipoles will cancel each other in most cases due to the
symmetry of the solid. However, if the crystal has reduced symmetry, the lower number
of symmetry operations that are allowed could lead to an only partial cancellation of
the internal dipolar fields, leaving the system electrically polar. Ferroelectrics are
pyroelectrics in which the direction of spontaneous polarization can be changed (for
example, rotated by 180) by applying a mechanical deformation or an external
electric field. All these phenomena find a great number of applications in modern
technology, from sensors and actuators, to memory cells and spintronics devices. In
general, pyro- and piezoelectric materials to be used in modern technological appli-
cations should display an excellent piezoelectric response, combined with high
mechanical stability and low environmental impact. Existing materials, which can
be broadly divided into the families of perovskite crystals, ceramics, and polymers,
can only partially fulfill the aforementioned requirements. Lead zirconate titanate
(PZT) ceramics materials, for example, are strong piezo- and pyroelectrics
1,2
with
spontaneous polarizations of 0.3 to 0.7 C/m
2
(Coulomb/m
2
), but unfortunately they
are also brittle, heavy, and toxic. On the other hand, polymers such as polyvinylidene
fluoride (PVDF) are lightweight, flexible, and virtually inert, but display polar
properties an order of magnitude weaker than those of PZT.
3
In Table 21.1 we
summarize the main characteristics of modern piezo- and pyroelectric materials,
compared with the predicted polar properties of BN nanotubes, which we will discuss
at length in this chapter.
In this context, carbon nanotubes and related nanotubular structures can play
an important role in advancing the field toward new and improved materials for
polarization-based applications. After their discovery,
4
it was almost immediately
shown that carbon nanotubes possess outstanding potential for technological appli-
cations due to their mechanical and electrical properties.
5
The advent of carbon
nanotubes, together with recent progress in nanomaterials design and processing,
has led to a quest for other novel graphene-based materials with technologically
desirable properties. Furthermore the closely related boron nitride nanotubes
(BNNTs) and mixed BN-C systems,
611
now produced in macroscopic quantities,
have electronic properties that are complementary to pure carbon nanotubes and
could therefore be useful in a variety of novel electronic devices. For instance, an
early theoretical study predicted that BN-C junctions may well be a practical way
to realize stable, nanoscale heterojunctions,
12
as exemplified in Fig. 21.1.
BNNTs, extensively investigated since their initial prediction
6
and succeeding
synthesis,
7
are well known for their excellent mechanical properties.
13
However,
unlike carbon nanotubes, the majority of BN structures are noncentrosymmetric and
polar, a fact that suggests the existence of nonzero spontaneous polarization fields.
T
A
B
L
E

2
1
.
1
S
u
m
m
a
r
y

o
f

P
r
o
p
e
r
t
i
e
s

o
f

P
o
l
a
r

M
a
t
e
r
i
a
l
s

a
n
d

B
o
r
o
n

N
i
t
r
i
d
e

N
a
n
o
t
u
b
e
s

M
a
t
e
r
i
a
l

C
l
a
s
s
R
e
p
r
e
s
e
n
t
a
t
i
v
e
s
P
r
o
p
e
r
t
i
e
s

(
C
/
m
2
)
P
r
o
s
C
o
n
s
P
o
l
a
r
i
z
a
t
i
o
n
P
i
e
z
o
e
l
e
c
t
r
i
c
i
t
y
L
e
a
d

z
i
r
c
o
n
a
t
e

t
i
t
a
n
a
t
e

(
P
Z
T
)

c
e
r
a
m
i
c
s
P
b
T
i
O
3
P
b
Z
r
O
3
P
b
Z
r
x
T
i
1
x
O
3
U
p

t
o

0
.
9
5
1
0
G
o
o
d

p
y
r
o
-

a
n
d

p
i
e
z
o
e
l
e
c
t
r
i
c

p
r
o
p
e
r
t
i
e
s
H
e
a
v
y
,

b
r
i
t
t
l
e
,

t
o
x
i
c
P
o
l
y
m
e
r
s
P
o
l
y
v
i
n
y
l
i
d
e
n
e

f
l
u
o
r
i
d
e

(
P
V
D
F
)
,
P
V
D
F

c
o
p
o
l
y
m
e
r
s
0
.
1
0
.
2
U
p

t
o

0
.
2
L
i
g
h
t
,

f
l
e
x
i
b
l
e
P
y
r
o
-

a
n
d

p
i
e
z
o
e
l
e
c
t
r
i
c

p
r
o
p
e
r
t
i
e
s

w
e
a
k
e
r

t
h
a
n

i
n

P
Z
T

c
e
r
a
m
i
c
s
B
o
r
o
n

n
i
t
r
i
d
e

n
a
n
o
t
u
b
e
s
Z
i
g
z
a
g

B
N

n
a
n
o
t
u
b
e
s
S
i
n
g
l
e

N
T
:
0
B
u
n
d
l
e
:
~
0
.
0
1
S
i
n
g
l
e

N
T
:
0
.
2
5
0
.
4
B
u
n
d
l
e
:
?
L
i
g
h
t
,

f
l
e
x
i
b
l
e
;

g
o
o
d

p
i
e
z
o
e
l
e
c
t
r
i
c

p
r
o
p
e
r
t
i
e
s
;

c
o
u
l
d

b
e

u
s
e
d

i
n

n
a
n
o
d
e
v
i
c
e
s
E
x
p
e
n
s
i
v
e
?
C
o
p
y
r
i
g
h
t

2
0
0
6

b
y

T
a
y
l
o
r

&

F
r
a
n
c
i
s
Recently these properties have been partially explored by Mele and Krl, using a
model electronic Hamiltonian,
14
which predicted that BNNTs are piezo- and pyro-
electric materials whose electronic polarization orientation depends on the geomet-
rical structures of the tubes.
In this chapter, we will review what we have learned in the past few years about
spontaneous polarization and piezoelectricity in BNNTs and composite BN-C nano-
tube structures.
15,16
We will estimate their potential usefulness in various pyro- and
piezoelectric device applications, and we will elucidate the interplay between sym-
metry and polarization in nanotubular systems. Finally we will discuss some recent
experimental results and relate them to our theoretical understanding.
In Section 21.2 we briefly show how modern polarization theory is formulated
employing the Berry phase or the Wannier functions approaches and present the
details of the numerical techniques we have employed to compute polarizations in
Section 21.3. (Subsections 21.2.1, 21.2.2, and 21.2.3 are quite technical and could
be skipped during the first reading.) In Section 21.4 we will discuss the results of
our computations of spontaneous polari-zation and piezoelectricity in BNNTs uti-
lizing both of the prescriptions given in Section 21.2. Section 21.5 will be devoted
to the analysis of the piezoelectric properties of these systems, while in Section 21.6
we will discuss a few instances where polarization effects can be indeed observed
in nanotubular structures. Conclusions and further discussion will end the chapter.
21.2 MODERN THEORY OF POLARIZATION
Macroscopic polarization of a dielectric is a central concept of classical electrostat-
ics, well known from elementary introductory physics classes. At a first glance, it
might seem that it is a well-defined quantity easily obtainable via simple analysis
or calculations. Most introductory condensed matter physics books contain a straight-
forward recipe to compute polarization in a solid that looks like this:
(21.1)
FIGURE 21.1 Electronic density in a nanotube containing a section of C (left) and a section
of BN (right). Note the difference in bonding, from quasiplanar sp
2
in C to ionic in BN.
P b r r r +
,
,
,
]
]
]
]

1
V
eZ d
l l
l
( )
Here the first term accounts for the polarization associated with ions of charge Z
l
positioned at b
l
and the second term represents the electronic contribution to
polarization, where U(r) is the density of electronic charge. If the summation and
integration in Equation 21.1 are extended over the whole macroscopic sample, i.e.,
index l runs over all ions and U(r) contains information about all electronic charges
including the ones on the crystal boundary, then the sum of the two terms is
absolutely convergent and the polarization P is well defined. However, in real
calculations it is usually impossible to sum and integrate over the whole sample.
Instead a unit cell representative of the characteristic features of the system under
investigation is chosen and periodically repeated over the whole space. In other
words, a finite piece of solid with a boundary is substituted with an infinite and
periodic solid without a boundary. For the latter, most condensed matter textbooks
introduce a dipole moment associated with the unit cell by limiting the summation
and integration in Equation 21.1 to the atoms and electrons within the cell. In this
case, index l runs over all ions in the cell and U(r) is the cell-periodic density of
electronic charge. The total polarization in the sample is then computed as a sum
over the unit cell dipole moments. There is a serious problem with this definition
since the aforementioned sum is now only conditionally convergent, i.e., it can
converge to a number of different values based on the choice of a given order of
summation. This means that with the exception of a very special case, when the
solid can be clearly partitioned into a set of charge neutral, weakly interacting
entities (the so-called Clausius-Mossotti limit), polarization becomes an ill-defined
quantity. Of course, the Clausius-Mossotti limit cannot be applied to a covalent solid
where electrons are delocalized.
One of the most relevant results in the modern theory of polarization is that
macroscopic polarization of a material cannot be predicted on the sole basis of its
microscopic charge distribution. However, it is obvious that it should also be a bulk
property of the material, i.e., independent of the shape of the sample or the choice
of the unit cell. The modern theory states that polarization in a solid can be defined
analogously to actual measurement procedures used in experiments. In reality, such
experiments never measure an absolute value of polarization, but rather a difference
P between two states of the material. This difference is indeed an intrinsic property
and does not depend on the truncation of the sample. The great achievement of
modern polarization theory is the correct definition and an operative way to compute
polarization differences in materials. In what follows, we will briefly outline two
different, although equivalent, ways to compute polarization in materials and nano-
structures: the Berry phase method and the Wannier functions method. For a com-
prehensive formal treatment of modern polarization theory, we refer the reader to
the papers by D. Vanderbilt and R. Resta.
17,18
21.2.1 COMPUTING POLARIZATION USING BERRY PHASE METHOD
The problem of computing polarization in materials is very subtle and can be
approached by the so-called Berry phase method, introduced only a decade ago by
Vanderbilt and King-Smith
17
and Resta.
18
Although a full description of the method
is beyond the scope of this chapter, it is instructive to outline the main ideas behind
this approach. In the modern theory of polarization, the polarization difference
between two states of a material is computed as a geometrical quantum phase. In
practice, this difference, , can be obtained if one can find an
adiabatic transformation O, which connects the states O
1
and O
0
, while leaving system
insulating. In the spirit of Reference 19, P
(O)
is conveniently split into two parts,
and , corresponding to the ionic and electronic contributions, respectively.
In the case of paired electron spins, the expression for the total polarization of the
system can be written as follows:
(21.2)
where V is the volume of the unit cell, and are the charge and position of
the Wth atom in the cell, and is the occupied cell-periodic Bloch state of the
system. For the electronic part, an electronic phase (Berry phase) defined mod-
ulo 2S (that is, an uncertainty amounting to an integer number of 2S in the deter-
mination of ) can be introduced as
(21.3)
where G
D
is the reciprocal lattice vector in the direction D along which we would like
to compute polarization (for example, along the axis of a nanotube). Of course, it is
hard to determine from Equation 21.3 that is truly an angular variable-defined
modulo 2S. However, if we substitute from Equation 21.2 into 21.3 and do all
the algebra (see Reference 17 for details), we can prove that it is indeed the case.
In a similar fashion, one can construct an angular variable for the ionic part (which
is just a sum over the point charges), called in what follows the ionic phase, so that
the total geometrical phase is
(21.4)
The total polarization in the direction D becomes
where R
D
is the real-space lattice vector corresponding to G
D
, i.e., R
D
G
D
1.
21.2.2 POLARIZATION FROM WANNIER FUNCTIONS
Alternatively the electronic polarization of a system can be expressed in terms of
the centers of charge of the Wannier functions (WFs) of its occupied bands:
17,18
P P P
( ) ( )
1 0
P
ion
( )
P
el
( )
P P P b k
( ) ( ) ( ) ( ) ( )
ion el
e
V
Z
ie
d
2
8
3
uu u
i i
BZ
i occ
k k k
( ) ( )
( )
b
( )
u
i k
( )
( )
( )
( ) ( )
V e
el
G P
( )
P
el
( )
( ) ( ) ( ) ( )
+
Z G b
P R
( ) ( )
/ e V
P r r r
R R el i
i
i i
i
e
V
W W d
e
V
( ) ( ) ( ) ( )

2 2
where is the WF for band i, and is the corresponding center. As a
reminder for the readers, a Wannier function, labeled by the Bravais lattice vector
R and the band index i, is usually dened via a unitary transformation of the Bloch
functions
ik
(r) of the ith band:
where V is the volume of the unit cell and the integration is performed over the
entire Brillouin zone (BZ). It is easy to show that the WFs, as dened previously,
form an orthonormal basis set, and that any two of them, for a given index i and
different R and R
0
, are only translated images of each other.
The advantage of this formula is that it allows the recasting of our problem from
the summation over distributed charges to a summation over localized WF charge
centers, which corresponds to the Clausius-Mossotti limit when the polarization is
well dened. The localized WFs can be computed by a number of different tech-
niques, for example, by the direct minimization of the spread functional as the sum
of the second moments of the WFs corresponding to a particular choice of transla-
tional lattice vector:
20
.
In either case (Berry phase or WFs), due to the arbitrariness in the choice of the
phases of the Bloch functions, can be obtained only modulo 2eR/V. However,
the difference in polarization P remains well dened provided that |P
el
| << |2eR/V|.
The same indetermination issues apply to .
19
21.2.3 A SIMPLE WAY TO COMPUTE BERRY PHASES ON A COMPUTER
The main ingredient that has to be computed in order to obtain the Berry phase
is the following matrix element (see Equation 21.2): If in the
express-ion for in Equation 21.2 we split the 3-D integration over the Brillouin
Any crystal can be viewed as a lattice, a periodic set of points in space that obey a number of symmetry
operations. In particular, there exist only 14 different lattice types, the so-called Bravais lattices, each
characterized by a set of translational vectors R that can be used to generate lattice sites. In the solid,
atoms or groups of atoms are associated with each lattice site, resulting in a denite crystal structure.
The periodicity of the solid also determines basic symmetry properties of the quantum mechanical
behavior of the electrons. Electrons in solids are described by wave functions (Bloch states) that have
the same periodicity as the crystalline lattice and an energy spectrum that reects the symmetric properties
of the crystal. These are usually characterized in terms of reciprocal lattice vectors G, the Fourier transforms
of the translational vectors R. Reciprocal lattice vectors dene the reciprocal of the crystal lattice, the
so-called Brillouin zone. As the Bravais lattice reects the symmetry of the structure of the solid, the
Brillouin zone reects the symmetry of the electronic states. For more on the geometrical and electronic
structures of solids see, for instance, N.W. Ashcroft and N.D. Mermin, Solid State Physics, Saunders
College, Philadelphia, 1976.
W
iR
r
( )
( )
r
i
( )
W
V
e d
i i
i
BZ
R k
k R
r r k ( )
( )
( ) =

2
3

=
( )
W W W W
i i
j
i i
j
j
0 0 0 0
r
2
2
r
P
el
( )
P
ion
( )
( )
u u
i i k k k
( ) ( )
| | .

P
el
( )
zone into the linear integral along G
D
and an area integral in the perpendicular
direction, we can rewrite Equation 21.3 as
, (21.5)
where
.
The matrix elements can be computed by using finite differences
on a fine 1-D mesh (or a string) of k-points {k
j
} along G
D
, as shown in the upper panel
of Fig. 21.2. An expression for such a string phase can be written in the following
fashion:
. (21.6)
FIGURE 21.2 Partitioning of the Brillouin zone for the computation of the Berry phase
(example for the cell of orthorhombic symmetry). G
D
is the reciprocal lattice vector in the
direction in which we compute polarization. A is a 2-D slice of the Brillouin zone, perpendicular
to the direction of G
D
. A 2-D mesh of k-points {k
A
} is chosen in A. Then a fine 1-D mesh of
k-points {k
j
} in the direction of G
D
is introduced for each point k
A
.
( ) ( )
( )
( )

2
2
3
VG
d
A
k k
( ) ( ) ( )
( ) k
i dk u
k
u
nk
G
n
nk
0
occ

u u
nk
k
nk
( ) ( )
| |

( ) ( ) (
( ) lim ( ) limImln det k k

J
J
J
nk
u
j
) ( )
u
mk
j
J
j +
j
(
,
,
\
,
(
(
1
0
1
G
a
A
A
k
k
k
j +1
k
j
The two sets of wave functions required to obtain each of the matrix elements in
Equation 21.6 are computed in the neighboring k-points k
j
and k
j1
along the string,
also shown in the upper panel of Fig. 21.2. In the limiting case when k
j
k
J
, i.e.,
we are positioned at the top of the string and the following condition should be
satisfied for the Bloch-state sets at k-points k
0
and k
J
:
.
Finally we turn the 2-D integral in Equation 21.5 into a sum over a set of k-points
{k
A
`, shown in the lower panel in Fig. 21.2, obtaining the operational expression
for computing the Berry phase as an average of the string phases over the
area perpendicular to G
D
:
.
As we can see, in order to obtain the Berry phase we need to know all the
occupied Bloch states in every point of the 3-D k-point mesh {k
j
, k
A
} and
also compute the overlap matrix elements for the neighboring k-points
within the same string. The electronic part of the polarization is then recovered
from the Berry phase with the help of the following formula:
21.3 COMPUTATIONAL DETAILS
The starting point for our procedure is the first principles calculations of the elec-
tronic structures of the systems. We adopt a standard electronic structure method
based on self-consistent total energy and force minimization, which allows for
optimizing simultaneously the atomic positions and the corresponding electronic
wave functions. The electronic structure is described within the density functional
theory (DFT).
21
We used a multigrid-based total energy method, employing a real-
space grid as a basis,
22
as well as a more standard plane wave-based approach for
all the polarization calculations.
23
The internal consistency of the results obtained
using these two methods has been carefully checked. The results were obtained in
the local density approximation (LDA),
24
and although more sophisticated exchange
correlation functionals (e.g., generalized gradient approximations) can obviously be
used, they do not alter the major conclusions of this investigation. The electronion
interaction is described via norm-conserving pseudopotentials
25
in the form of Klein-
man and Bylander.
26
To isolate the contributions of individual nanotubes, we performed polarization
calculations for periodic crystals of noninteracting (i.e., positioned sufficiently far
apart) nanotubes (Fig. 21.3(a)). In order to assess the importance of close-packing,
the calculations were repeated for nanotubes arranged in a hexagonal lattice as
depicted in Fig. 21.3(b). The electronic structure calculations were carried out using
u e u
nk
i
nk
J
( ) ( )
( ) ( )

r r
G r

J
( )
( ) k
( ) ( )
( )

2
N
J
k
k
k
( )
u
nk
( )
( )
r

+
u u
n m
j
k k
j
( ) ( )
|

1
P R
el
e V
( ) ( )
/ .
( )

two special k-points along the *-A direction in the hexagonal or *-Z direction in
the tetragonal Brillouin zone. The k-space integration to compute was done on
a string of 20 k-points uniformly distributed along the same direction and shifted to
avoid the *-point.
21.4 POLARIZATION IN NANOTUBES
Calculations of spontaneous polarization in BNNTs with the techniques described in
the previous section require a reference state of the system that is intrinsically nonpolar.
The polarization difference 'P between the polar and the nonpolar states of the system
will then give the actual value of spontaneous polarization. An obvious choice for the
nonpolar state of a BNNT is a nanotube of the same geometry, but with boron and
nitrogen atoms substituted by carbon atoms. Such nanotubes are naturally nonpolar
because of the purely covalent bonds connecting all the atoms. However, we might
also want to compute the polarization difference in many points between the two
limiting cases of BNNT and CNT to make sure that each such difference remains
smaller than the polarization quantum 2eR/V. This can be done by employing a so-
called virtual crystal procedure where carbon atoms in the CNT are substituted by
pseudocarbon atoms, which are 50% boron and 50% nitrogen. The adiabatic transfor-
mation parameter O corresponds to the content (in atomic %) of a site that is gradually
transformed from pseudocarbon (O 50%) to pure boron or nitrogen (O 100%), as
shown in Fig. 21.4. Because of the different alignments of the polar bond with respect
to the nanotube axis, we anticipate that the symmetry of the nanotube will play an
important role in determining the magnitude of the spontaneous polarization field. In
particular, since the zigzag geometry maximizes the axial dipole moment, we expect
to observe the strongest effects in (n,0) nanotubes.
21.4.1 BERRY PHASE METHOD
The ionic part of the polarization in zigzag BNNTs presented in Fig. 21.5 is large
and directly proportional to the nanotubes index. This is in contrast, for instance,
with the corresponding wurtzite IIIV and IIVI systems,
27,28
where the spontaneous
FIGURE 21.3 Examples of two nanotube arrangements used in the simulations: (a) array of
nanotubes sufficiently far apart that can be considered noninteracting; (b) geometry of a unit
cell containing interacting nanotubes (rope).
(a) (b)

z
( )
polarization can be viewed as the difference between the polarizations of the
wurtzite (polar) and zinc blende (nonpolar) geometries. Since these configurations
become geometrically distinct only in the second shell of neighbors, their ionic
phases are very close. The major contribution to the spontaneous polarization in
wurtzite materials is then due to the difference between the electronic polarizations
(which are 0.04 to 0.08 C/m
2
), while in BNNTs both the ionic and the electronic
contributions are essential.
The ionic phase differences M
ion
between the polar and nonpolar configurations
of zigzag nanotubes were evaluated via the virtual crystal approximation. The inset
in Fig. 21.5 shows the results obtained by a simple lattice summation over the ionic
charges (the first term in Equation 21.2), with the phases translated into the [S, S]
interval. The phases plotted in the main graph were unfolded by eliminating all the
2S discontinuities and setting the phase of the nonpolar reference configuration to
zero. For the unfolded phases, as the diameter of a nanotube increases, i.e., as another
hexagon is added around the circumference of the tube, the ionic phase increases
by S/3, so that the total ionic phase for a (n,0) BN nanotube amounts to nS/3.
In Fig. 21.6 we show the electronic phase differences M
el
between the polar and
nonpolar configurations for zigzag nanotubes. These data suggest a natural division
FIGURE 21.4 Virtual crystal procedure. Going from A to D, the B and N atoms are gradually
transformed into C, according to the variation of the parameter O.
of the nanotubes into three families with different M
el
: S /3 for n 3l 1, S / 3
for n 3l 1, and S for n 3l, where l is an integer, which is similar to the
result obtained by Mele and Krl.
14
It is important to note that in the limiting case
of flat C or BN sheets, the electronic polarization is zero for any value of O, due to
the existence of a three-fold symmetry axis perpendicular to the surface of the sheet.
The existence of the three classes of behavior in BN nanotubes is surprising, given
that the ionic character of the electronic charge density (associated with the BN
FIGURE 21.5 Ionic phase (see Equation 21.4) differences between the polar and nonpolar
configurations for zigzag nanotubes; the ionic phase of the nonpolar configuration is set to
zero. Inset: ionic phases wrapped into the [S, S] interval. Phases are given in units of S.
FIGURE 21.6 Electronic phase (see Equation 21.3) differences between the polar and non
polar configurations for zigzag nanotubes.
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
50
50 60 70 80 90 100
60 70 80 90 100
(atomic %)
(atomic %)
Supercell with
three hexagons
5.0
6.0
7.0
8.0
9.0
10.0
Sheet
10/3
9/3
8/3
7/3
6/3
5/3
3/3
i
o
n

(
i
o
n
(
)
1
0
1
50 60 70 80 90 100
(atomic %)
e
l

(
)
1.0
0.5
0.0
0.5
1.0
5.0
6.0
7.0
8.0
9.0
10.0
Sheet
bond) does not change with the nanotube index. Additionally, there is an important
difference between our results and those of Reference 14, where electronic polar-
ization of heteropolar nanotubes was studied within a simple S-orbital tight-binding
(S-TB) approximation. In Reference 14 the n 3l family has a zero electronic phase
instead of S.
This discrepancy is due to the ambiguity of the definition of electronic polariza-
tion as a multivalued quantity,
19
which can assume a lattice of values corresponding
to Berry phases that differ by arbitrary multiples of 2S. Unlike the ionic phase model,
where discontinuities in can be easily monitored, Berry phase calculations
always produce phases that are smoothly folded into the [S, S] interval and cannot
be extrapolated. To obtain an unambiguous determination of the spontaneous polar-
ization of BNNTs of arbitrary diameters, one has to compute the polarization in a
different way, using the centers of charge of the WFs of the occupied bands. Note
that this approach does not solve the problem of branch indetermination, since as
Berry phases are defined modulo 2S, Wannier centers are defined modulo a lattice
vector R.
#
However, by shifting the indetermination from the phase to the lattice
vector, we are able to map the electronic polarization problem onto a simple elec-
trostatic model, where the unfolding of the electronic phase is straightforward.
19
21.4.2 MAXIMALLY LOCALIZED WANNIER FUNCTIONS
The results of the maximally localized WF calculations for BNNTs are summarized
in Fig. 21.7, where examples of the WFs for C and BN zigzag nanotubes of arbitrary
diameter are shown, together with a schematic drawing that illustrates the shift of
the Wannier centers in the adiabatic transformation from C to BN. Since
the magnitude of the shift of the centers is directly proportional to the electronic
polarization of the BNNT with respect to the nonpolar CNT.
The V-band WFs are centered in the middle of the CC bonds in carbon
nanotubes while in BNNTs they are shifted toward the cations because of the
different electronegativities of B and N atoms. Since these shifts have the same
magnitude along each of the three bond directions, the vector sum of all shifts is
zero (see bottom panel of Fig. 21.7), and the V orbitals do not contribute to the total
polarization of the system. The S-band WFs are centered on the cations in BNNTs,
while in CNTs they have a peculiar V-shape, with centers somewhat outside of the
CC bond. The sum of the shifts of the S-band Wannier centers is nonzero only for
the axial component, which means that the electronic polarization in BNNTs is
purely axial.
The bottom panel of Fig. 21.7 shows the projection of the S WF centers onto
the axis of the tube. The projections of the centers have an effective periodicity of
half of the axial lattice constant c, which leads to the indetermination of the electronic
#
Wannier functions and their centers have the periodicity of the lattice.

ion
( )
P r r
el
BN
i
BN
i
C
i
e
V

( )
2
phase by multiples of S. Moreover the WF description allows for an unambiguous
unfolding of the electronic phase. In analogy to the ionic phase, we find that each
individual hexagon carries a phase of S/ 3, leading to a total electronic phase of
nS/ 3 for a (n,0) nanotube. This result demonstrates that the direction of the
electronic polarization in a BNNT is specified by the orientation of the BN bond
and does not oscillate in direction with the nanotube diameter, contrary to the model
Hamiltonian predictions.
14
We should point out that the Wannier function results
are completely consistent with the Berry phase calculations, since an electronic
phase of nS / 3 for any n can be folded modulo S into the three families found
previously.
When we combine the results for the ionic and electronic phases into a general
formula for the phase of any (n,m) BNNT (n and m being any integer number), we
find that the two contributions cancel exactly and that the total spontaneous polar-
ization in any BNNT remains zero, i.e., the Wannier centers are positioned in such
a way as to completely compensate the polarization due to ions. This result was
confirmed by two-point (corresponding to O 50 and 100%) calculations of the
Berry phase difference for a number of chiral nanotubes ((3,1), (3,2), (4,1), (4,2),
(5,2), and (8,2)) and the exact cancellation of the polarization contribution was
obtained for all BNNTs, except for those having diameters smaller than about 4 ,
where a residual polarization subsists as the consequence of the very high curvature.
In such nanotubes, Wannier centers cannot fully compensate the ionic polarization
because of the severe distortion of the atomic bonds, which makes these systems
FIGURE 21.7 Upper panel: examples of Wannier functions (WFs) of the V and S occupied
bands of C (left panel) and BN (right panel) nanotubes. Lower panel: schematic positions of
the centers of the Wannier functions in C and BN hexagons, and the projections of the S WFs
onto the nanotube axes. The positions of the centers of V WF are indicated by circles, and
those of S by squares. The direction of the shifts of V and S WFs in an adiabatic transformation
from C to BN is indicated by arrows. The projections of shifts of the V WFs cancel, so that
the V WFs do not contribute to polarization (see text).
C
0 0 1c 1/2c 1/2c 1c
B
BN
N
weakly pyroelectric. For example, we found that P
3
0.11 C/m
2
in (3,1), 0.008 C/m
2
in (7,0), and 0.002 C/m
2
in (12,0) nanotubes.
The exact cancellation is a result of the overall chiral symmetry of the nanotubes,
which, although not centrosymmetric, are intrinsically nonpolar. However, cancellation
of ionic and electronic polarizations is exact only in the limit of an isolated BNNT.
The spontaneous polarization in a nanotube bundle, where the chiral symmetry is
effectively broken, differs from zero. For example, in a bundle of (7,0) tubes posi-
tioned at a wall-to-wall equilibrium distance of 3.2 , the polarization amounts to
P
3
| 0.01 C/m
2
. It must be pointed out that in this example it is difficult to estimate
the relative contributions to the polarization of the bundling, extreme curvature and
elastic deformation of the nanotubes. Although smaller than in polymers or PZT,
this polarization is comparable to some wurtzite pyroelectrics: e.g., P
3
0.06 C/m
2
in w-ZnO.
28
21.5 PIEZOELECTRICITY IN NANOTUBES
The Berry phase method can also be employed to compute piezoelectric properties
of BNNTs that are directly related to the polarization differences between strained
and unstrained tube geometries. In the linear regime, the change in polarization due
to strain can be decomposed into a sum of two terms: a uniform axial strain and a
relative displacement of the two sublattices. It is therefore natural to describe the
geometry of a BNNT of a given radius in terms of an axial lattice constant c and
an internal parameter u, uc therefore being the length of the vector connecting the
anion to the cation. With this choice, the axial piezoelectric polarization is
(21.7)
where the strain H
3
(c c
0
)/c
0
, and c
0
and u
0
are the equilibrium values of c and
u. Then the only surviving piezoelectric stress tensor component is
where N is the number of BN pairs in the supercell. Here
is the so-called clamped-ion piezoelectric constant representing the effect of strain
on the electronic structure, and
P e
P
c
c c
P
u
u u
3 33 3
3
0
3
0

( ) ( )
e e
ec
V
NZ
du
dc
33 33
0
0
2
+
( ) *
e c
P
c
33
0
0
3
( )
Z
V
eNc
P
u
*
0
3
is the axial component of the Born dynamical charge tensor. Both polarization
derivatives were computed as finite differences, changing c or u by r1%. The
parameter [ c
0
du/dc, which is a measure of the bond length change under axial
strain, was obtained by rescaling c (together with the associated components of
ionic coordinates) and then relaxing the geometry of the system. For all the
systems considered subsequently, the value of [ is approximately the same and
equal to 0.085.
We have calculated the piezoelectric properties for various bundles composed
of zigzag BNNTs with individual diameters ranging from 3.9 to 10.2 . These results
are summarized in Table 21.2 and compared to a few well-known piezo- and pyro-
electric materials. While the piezoelectric constants of zigzag BNNTs are modest
when compared with inorganic compounds, they are still substantially larger than
those in the PVDF polymer family. In addition, we should point out that piezoelectric
properties of multiwall BNNTs, where all individual nanotube layers have the same
orientation of the BN bond, would be greatly magnified (the total piezoelectric
constant will be equal to or larger than the sum of the constants of the individual
tubes composing the multiwall BNNT) and could reach the same order of magnitude
as in ferroelectric ceramics. In a recent paper by Sai and Mele
29
both uniaxial and
shear piezoelectric distortions were considered in heteropolar nanotubes and a model
connecting piezoelectric properties of nanotubes with the limiting case of the flat
sheet was constructed, allowing for a simple derivation of piezoelectric constants in
chiral nanotubes.
TABLE 21.2
Piezoelectric Properties of Zigzag BNNT Bundles
(n, m) Diameter () Z
*
|e
33
| (C/m
2
) Ref.
(5,0) 3.91 2.739 0.389
(6,0) 4.69 2.696 0.332
(7,0) 5.47 2.655 0.293
(8,0) 6.24 2.639 0.263
(9,0) 7.04 2.634 0.239
(10,0) 7.83 2.626 0.224
(11,0) 8.57 2.614 0.211
(12,0) 9.38 2.609 0.198
(13,0) 10.26 2.605 0.186
w-AlN 2.653 1.50 [27]
w-ZnO 2.110 0.89 [28]
PbTiO
3
3.23 [2]
P(VDF/TrFE) |0.12 [3]
The corresponding values for a few piezoelectric materials are listed for comparison. The data
reported in the table are for nanotube bundles where we assume a hexagonal close-packed geometry
with intertube equilibrium distance of 3.2 .
In addition to the piezoelectric stress constants e
DE
connecting polarization and
strain (see Equation 21.7), it is informative to take a look at piezoelectric strain
constants d
DE
that couple polarization and stress or, conversely strain and electric
field.
30
Knowledge of these constants (again, only the d
33
constant is important due
to the reduced dimensionality of the system) would provide some insight about
possible application areas for BNNTs. Piezoelectric strain constant d
33
can be
obtained from e
33
by dividing it by the tubes Youngs modulus. By using Table 21.2
and a typical value of 0.9 TPa for the Youngs modulus,
31
we can estimate d
33
as
approximately 0.3 pC/N. This allows us to evaluate actuating abilities of zigzag BN
nanotubes. For example, to obtain a 3% elongation (strain) of the tube, an electric
field in the order of 100 V/nm has to be applied. This leads to the conclusion that
BN nanotubes will probably be much better sensors (high piezoelectric stress con-
stants) than actuators (low piezoelectric strain constants). In ferroelectric polymers
such as PVDFs the situation is inverse. Their piezoelectric strain constants are two
orders of magnitude larger than in BN nanotubes (30 pC/N is a typical value) and
piezoelectric stress constants are low (about 0.1 C/m
2
).
21.6 POLARIZATION EFFECTS IN NANOTUBULAR
STRUCTURES
Although they have zero macroscopic polarization when isolated, and they are only
mildly pyroelectric when bundled, BN nanotubes, when properly used in complex
systems, can still be viewed as intrinsically polar materials and their properties can
therefore be exploited toward technological applications. They are only intrinsically
nonpolar in the ideal infinitely long configuration, but nonvanishing polar properties
should manifest themselves when they are part of more complex designs, such as
nanotube superlattices and heterostructures or, in general, nanostructures with sym-
metry properties reduced with respect to the cylindrical geometry of the nanotubes.
In general, superlattices alternating polar and nonpolar materials should electroni-
cally respond to the internal electric fields, with phenomena such as charge separa-
tion and screening, or as we will see, enhanced field emission properties in finite,
tip-shaped samples. Incidentally our calculations on BN-C heterojunctions and BN-
C coaxial systems show that the spontaneous polarization fields associated with B
and N atoms do indeed dramatically enhance field emission properties and lead to
attractive electronic devices.
Carbon nanotubes are generally considered to be good electron emitters, owing
to their electronic properties and exceptional aspect ratios. The remarkable emitting
properties may be further improved by exploiting the electronic properties of BN-
doped carbon systems. The idea here is to make use of the intrinsic polarization
field as means to reduce the difficulty to extract electrons, i.e., the work function, at
the tip of the system, thereby enhancing the extraction of electrons from the system.
21.6.1 NANOTUBE SUPERLATTICES AND HETEROSTRUCTURES
The orderly introduction of BN into otherwise pure carbon nanotubes may be
realized in two different manners: either by substituting one CC pair with one BN
pair in the carbon nanotube lattice and therefore mixing the C and BN phases
(Fig. 21.8(a)), or by creating a system that presents two coaxial homogeneous phases
of C and BN (Fig. 21.8(b)). While early attempts to design BN-C heterojunctions
by self-assembly have been so far unsuccessful, and actually led to the creating of
biwalled BN-C systems, recent progress in the now well controlled substitution
reaction (in which CC pairs are substituted by BN pairs) offers a potentially new
route for the design of BN-C heterojunctions.
Since BN-C nanotube superlattices are quasi-one-dimensional (1-D) heterojunc-
tions, their electronic properties are first and foremost characterized by their band
offsets. In contrast to 3-D solids, the asymptotic band offsets of such heterojunctions
are conveniently computed from the bulk properties of their isolated components,
since the vacuum level may be used as unequivocal reference energy.
32
Pure BN
nanotubes are wide band gap semiconductors with band gaps of 5.5 eV. Carbon
nanotubes, on the other hand, may be either metallic or semiconducting, depending
on their chiral indices. The (l,0) zigzag tubes on which we will concentrate here are
either metallic (l 3n, where n is an integer) or semiconducting. In BN-C nanotube
systems, the valence and conduction band offsets are spatially direct, with typical
VLDA values given in Table 21.3. From this table, it is clear that simple matching
of two straight nanotubes will give rise to a variety of band alignments with band
offsets that are sensitive to the helicities and radii of the tubes. In particular, for
FIGURE 21.8 Prototypes of BN-C hybrids: (a) a BN-C heterojunction based on an (11,0)
nanotube and (b) a (9,0) C and (17,0) BN biwalled system.
TABLE 21.3
Conduction (E
c
) and Valence (E
v
) Band Offsets for a Set of Zigzag BN-C
Heterostructures Relative to Vacuum Level
(7,0) (8,0) (9,0) (10,0)
E
c
(eV) 2.09 2.50 2.42 2.52
E
v
(eV) 0.82 1.12 1.53 1.46
metallic carbon nanotubes, the heterojunction will behave as a Schottky diode. Thus
combining the BN and carbon nanotubes of various symmetries offers a broad range
of opportunities for band offset engineering and for the design of a number of
different nanotubular heterojunctions.
The majority of BN-C nanotube superlattices are characterized by an axial dipolar
electric field, which originates from the polar nature of the BN bond. In a periodic
superlattice of polar and nonpolar materials (such as BN and C), any polarization
field will affect the behavior of the total electrostatic potential of the system. In fact,
the latter displays a typical sawtooth behavior that is the signature of the presence of
the field superimposed onto the periodic crystalline potential.
33
For convenience, we
have computed the planar average of the electrostatic potential over the directions
perpendicular to the tube axis, as shown in Fig. 21.9(a). The average potential displays
strong oscillations due to the varying strengths of the ionic potentials. To eliminate
this effect, we implemented the procedure of Reference 34, and thereby calculated
the one-dimensional macroscopic average of the electrostatic potential of the system.
The value of the dipole field was then obtained from the slope of the macroscopic
FIGURE 21.9 (a) Average electrostatic potential (solid line) and macroscopic average (dotted
line) along a BN-C (6,0) superlattice (insert) and (b) macroscopic polarization field for zigzag
(l,0) tubes as a function of the helicity index. The asymptotic value for planar BN is represented
by the horizontal line.
0 5 10 15
Axial distance ()
0.0
2.0
4.0
6.0
8.0
P
o
t
e
n
t
i
a
l

(
e
V
)
(a)
0.1
0
0.1
0.2
0.3
0.4
0.5
2 4 6 8 10 12 14
Nanotube index
P
o
l
a
r
i
z
a
t
i
o
n

f
i
e
l
d

(
e
V
/
)
(b)
average potential and is shown in Fig. 21.9(b). Since the geometry of the zigzag
nanotubes maximizes the dipole moment of the BN bond, the strongest effects will
be observed in them. Hence we have primarily concentrated on zigzag nanotubes,
while considering the flat sheet system as the limiting case for large diameter tubes
(see Fig. 21.9(b)). In contrast, the (l,l) armchair nanotubes, in which one-third of
the bonds are perpendicular to the tube axis, are not expected to display any dipole
field. This is because any individual nanotube ring is charge neutral and therefore
no field can develop. Chiral nanotubes have fields that are between the zigzag and
armchair values and depend on the individual terminations. From Fig. 21.9(b),
we see that the magnitude of the macroscopic electric fields for the various super-
lattices has a strong oscillatory character. Nanotubes with helicity indices 7, 10, . .
. , 3n 1 are characterized by a substantially smaller field. This effect originates
from the fact that the carbon sections effectively screen the macroscopic polarization
field with an efficiency that depends on the nanotube helicity.
To understand that effect, one should analyze the electronic state at the valence
band edge of the BN-C superlattice, which as the BN-C band offset calculations have
shown, is always spatially localized in the carbon region.
16
This state plays an
important role in determining the response to the macroscopic electric field induced
by the BN section. For small-diameter nanotubes, the valence state displays either a
longitudinal or a transverse symmetry as exemplified by the (7,0) and (8,0) tubes on
Fig. 21.10.
35
Specifically for (l 3n 1,0) nanotubes, such as the (7,0) tube, the
valence state always assumes a longitudinal character. The axial distribution of valence
electrons therefore induces a depolarization field that is opposite to the one induced
by the BN section. This field is very effective in small-diameter nanotubes, so that
for the (7,0) nanotube the screening is almost total and the observed macroscopic field
is close to zero. For larger diameter nanotubes, the symmetry of the valence state
gradually loses its strong axial or longitudinal character and the macroscopic field
asymptotically approaches the value found for a flat BN sheet in the zigzag direction.
Despite the screening by the valence electrons, which are distributed over the
carbon portion of the BN-C superlattices, a net polarization field is built up along
any typical zigzag structure. The existence of such a macroscopic field clearly influ-
ences the extraction of electrons from BN-C systems. Qualitatively a good electron
emitter is characterized by a large geometrical field enhancement factor E and a small
work function I, the latter being an intrinsic property of the emitter material.
FIGURE 21.10 (Color figure follows p. 12.) Three-dimensional plots of the valence state
for (7,0) and (8,0) BN-C superlattices.
BN
(7.0) (8.0)
C
Quantitatively the current at the emitter tip is calculated from the well-known
FowlerNordheim relationship.
36
To assess the efficacy of BN-C systems as emitters, we have computed the field
enhancement factor and the work functions for tubes 15.62 in length in an applied
field of 0.11 V/. As expected, the applied field was very well screened inside the
system and a local field enhancement factor of 2.1 was obtained for the different
BN-C systems. This value is very close to the one obtained on the corresponding
pure carbon nanotube, which is not surprising since the field enhancement factor
typically depends only on the emitter geometry. Since the field enhancement factor
increases linearly with the sizes of the nanotubes, large enhancement factors are
clearly possible for long nanotubes.
37
Turning to the effect on the work functions,
we have built a set of finite-sized (6,0) zigzag structures using B, N, and C in various
combinations. The work functions were then computed as the difference between
the vacuum level and the Fermi energy of the system. The former was obtained
using the previously discussed potential average procedure, while the latter was
simply given by the highest occupied KohnSham eigenstate of the system. In order
to avoid any spurious effects arising from the periodic calculation, we have used the
so-called dipole correction in a unit cell surrounded by a vacuum of at least 15 .
38
Due to the net polarization field experienced by the electrons, the work function in
a NB-C heterostructure is reduced to 5.04 eV at the C tip and increased to 7.52 eV
at the N tip. The same trend is observed for the BN-C system, for which the work
function at the B tip is equal to 5.00 eV, while it takes a value of 6.45 eV at the C
tip. The work function is therefore decreased by a significant 1.40 eV, as compared
to the pure carbon system. This large difference leads to significant macroscopic
effects. According to the FowlerNordheim relationship, the logarithm of the current
density (J) depends upon the work function I as in
where E is the applied electric field, I is the work function at the tube tip, and E is
the field enhancement factor. It follows that the insertion of BN segments in C
nanotubes may increase the current density by as much as two orders of magnitude
as compared to pure carbon nanotube systems.
21.6.2 MULTIWALLED HYBRIDS
The previous discussion showed that the presence of BN atoms induces a significant
improvement of field emission properties at the carbon tip in a NB-C heterojunction.
However, the use of such devices as performant field emitters is limited by the major
issue related to their synthesis. From a practical point of view, the related nested BN-
C or C-BN systems offer a more direct way to take advantage of the polarity associated
with the BN phase. To assess the field emission properties of BN-C devices, we have
performed calculations of the field enhancement factor and of the work function. We
have specifically investigated the effect of the presence of a finite BN nanotube shell
surrounding a C nanotube: the C (5,5)-BN (18,0), as shown on Fig. 21.11. As expected
J
C
E e
B
E
2
3 2 /
from classical electrostatics, the field enhancement factor at the carbon tip is not
modified by the presence of the BN shell, since the insulating BN layer does not
disturb the outstanding screening properties of the underlying C tube. On the contrary,
the dipole field induced by the BN shell produces a net decrease of the work function
(by about 0.5 eV) and a subsequent increase of the emitted current (by up to a factor
of five), as compared to the current emitted by an isolated (5,5) carbon nanotube.
21.6.3 EXPERIMENTAL RESULTS
Although field emission properties of carbon nanotubes are at the forefront of applied
nanotechnology, a number of parameters for composite BNC and BN nanotubes remain
not well established and a direct connection between theoretical predictions and exper-
imental measurements cannot yet be made at the time of this writing. However, there
are already indications that given a better control of the growth and the structures of
these systems, their polarization properties might play a significant role in advanced
applications. CN
x
and BNC
x
nanotubes have already shown remarkable, up to 4 PA,
emission currents. This value exceeds by a factor of two the maximum stable field
emission current obtained from pure carbon nanotube samples.
39
The same excellent
field emission properties have been observed in aligned multiwalled nanotubes where
the outer layer is a BN-rich nanotube that insulates semiconducting BNC layers.
40
21.7 CONCLUSIONS AND FUTURE PROSPECTIVES
From the preceding discussion, it is clear that the engineering of nanostructures and
nanocomposites that incorporate BN nanotubes could potentially lead to advanced
device architectures for strain sensors and actuators, not to mention field effect
devices. Three main technological challenges that have to be overcome before any
useful devices with BN nanotubes can be produced are the following.
First, BN nanotubes have to be grown in quantity. Substantial progress has
been achieved in this area in recent years with numerous techniques for BN
FIGURE 21.11 Electron density 5 from the tip of a C (5,5)-BN (18,0) biwalled system.
nanotube synthesis such as, for example, plasma arc discharge, laser ablation/heat-
ing, substitution reaction from carbon nanotubes or various chemical methods
becoming available. It is noteworthy that BN nanotubes prefer to grow in zigzag
conformations,
41
which eliminates the problem of separating the tubes with different
chiralities, an issue that has not yet been completely solved for carbon nanotubes.
However, a different synthesis-related problem arises for BN nanotubes. In order for
multiwall tubes to be piezoelectric, they must be grown in such a way that all the
individual layers composing the multiwall tubes have the same direction as the polar
BN bond. A possible approach here could be to grow BN nanotubes in a strong
electric field so that only a certain BN bond direction would be preferred.
Second, efficient self-assembly procedures are required to make ropes or possibly
molecular crystalline structures from individual tubes, again preserving the same
direction of BN bond among them.
Finally, advanced nanomanipulation techniques should be developed to convert
pyro- and piezoelectric or field-emitting nanotube ropes and/or crystals into useful
devices. This includes, for example, attaching contacts or covering them with pro-
tective coatings.
Of course, the aforementioned techniques for incorporation of BN- and carbon-
based nanotubular structures into composite materials and their harnessing into useful
geometries for sensing or actuating purposes are still in their infancy and require further
investigation. However, we believe that the theoretical description of the microscopic
properties that characterize the polarity and piezoelectric responses of these systems
is a necessary and fundamental step for their possible industrial application.
PROBLEMS
Consider a hypothetical one-dimensional periodic (infinite) system, with lattice con-
stant R and with a unit cell composed of four atoms numbered 1 to 4 and positioned
relative to one another, as depicted in the figure. At equilibrium the system is such
that the net charge on atoms 1, 2, 3, and 4 is equal to Gq, Gq, Gq, and Gq, respectively.
1. Assuming that the ionic polarization is given by the dipole moment in a
unit cell per unit volume, explain the apparent discrepancy between the
values of the polarization computed using different choices of supposedly
equivalent unit cells A, B, C, and D defined on the figure.
2. Supposing that the system is truncated into a finite system, find the
optimal (in terms of maximum polarization) way to cut the system,
c
b
a
R
D
0
A
B
C
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
keeping the number of different atomic species equal. Show that the
charge surface actually plays a crucial role in defining the magnitude
of the polarization.
ACKNOWLEDGMENTS
The authors wish to thank Dr. Arrigo Calzolari for his help in computing polarization
properties from Wannier functions.
This research was sponsored in part by the Mathematical, Information and
Computational Sciences Division, Office of Advanced Scientific Computing
Research of the U.S. Department of Energy under Contract DE-AC05-00OR22725
with UT-Battelle, LLC.
REFERENCES
1. Kumazawa, T., Kumagai, Y., Miura, H., Kitano, M., and Kushida, K., Effect of external
stress on polarization in ferroelectric thin films, Appl. Phys. Lett., 72, 608, 1998.
2. Sghi-Szab, G., Cohen, R.E., and Krakauer, H., First-principles study of piezo-
electricity in PbTiO3, Phys. Rev. Lett., 80, 4321, 1998; First-principles study of
piezoelectricity in tetragonal PbTiO3 and PbZr1/2Ti1/2O3, Phys. Rev. B, 59, 12771,
1999.
3. Furukawa, T., Piezoelectricity and pyroelectricity in polymers, IEEE Trans. Electr.
Insul., 24, 375, 1989; Eberle, G., Schmidt, H., and Eisenmenger, W., Piezoelectric
polymer, IEEE Trans. Diel. Electr. Insul., 3, 624, 1996; Samara, G.A., Ferroelectricity
revisited advances in materials and physics, Solid State Phys., 56, 239, 2001.
5. Bernholc, J., Brenner, D., Buongiorno Nardelli, M., Meunier, V., and Roland, C.,
Mechanical and electrical properties of nanotubes, Annu. Rev. Mater. Res., 32, 347,
2002.
6. Rubio, A., Corkill, J.L., and Cohen, M.L., Theory of graphitic boron nitride nano-
tubes, Phys. Rev. B, 49, 5081, 1994.
7. Chopra, N.G., Luyken, R.J., Cherrey, K., Crespi, V.H., Cohen, M.L., Louie, S.G.,
and Zettl, A., Boron nitride nanotubes, Science, 269, 966, 1995.
8. Stephan, O., Ajayan, P.M., Colliex, C., Redlich, P., Lambert, J.M., Bernier, P., and
Lefin, P., Doping graphitic and carbon nanotube structures with boron and nitrogen,
Science, 266, 1683, 1994.
9. Loiseau, A., Willaime, F., Demoncy, N., Hug, G., and Pascard, H., Boron nitride
nanotubes with reduced numbers of layers synthesized by arc discharge, Phys. Rev.
Lett., 76, 4737, 1996.
10. Cumings, J. and Zettl, A., Mass production of boron nitride double-wall nanotubes
and nanococoons, Chem. Phys. Lett., 316, 211, 2000.
11. Lee, R.S., Gavillet, J., de la Chapelle, M.L., Loiseau, A., Cochon, J.-L., Pigache, D.,
and Willaime, J.T.F., Catalyst-free synthesis of boron nitride single-wall nanotubes
with a preferred zig-zag configuration, Phys. Rev. B, 64, 121405, 2001.
12. Blase, X., Charlier, J.C., DeVita, A., and Car, R., Theory of composite BxCyNz
nanotube heterojunctions, Appl. Phys. Lett., 70, 197, 1997.
13. Zhang, P. and Crespi, V.H., Plastic deformations of boronnitride nanotubes: an
unexpected weakness, Phys. Rev. B, 62, 11050, 2000; Bettinger, H.F., Dumitric, T.,
Scuseria, G.E., and Yakobson, B.I., Mechanically induced defects and strength of BN
nanotubes, Phys. Rev. B, 65, 041406, 2002.
14. Mele, E.J. and Krl, P., Electric polarization of heteropolar nanotubes as a geometric
phase, Phys. Rev. Lett., 88, 056803, 2002.
15. Nakhmanson, S.M., Calzolari, A., Meunier, V., Bernholc, J., and Buongiorno-
Nardelli, M., Spontaneous polarization and piezoelectricity in boron nitride nano-
tubes, Phys. Rev. B, 67, 235406, 2003.
16. Meunier, V., Roland, C., Bernholc, J., and Buongiorno Nardelli, M., Electronic and
field emission properties of boron nitride/carbon nanotube superlattices, Appl. Phys.
Lett., 81, 47, 2002.
17. King-Smith, R.D. and Vanderbilt, D., Theory of polarization of crystalline solids,
Phys. Rev. B, 47, 1651, 1993; Vanderbilt, D. and King-Smith, R.D., Electric polar-
ization as a bulk quantity and its relation to surface charge, Phys. Rev. B, 48, 4442,
1993.
18. Resta, R., Macroscopic polarization in crystalline dielectrics: the geometric phase
approach, Rev. Mod. Phys., 66, 899, 1994.
19. Vanderbilt, D., Berry phase theory of proper piezoelectric response, J. Phys. Chem.
Solids, 61, 147, 2000.
20. Marzari, N. and Vanderbilt, D., Maximally localized generalized Wannier functions
for composite energy bands, Phys. Rev. B, 56, 12847, 1997; Souza, I., Marzari, N.,
and Vanderbilt, D., Maximally localized Wannier functions for entangled energy
bands, Phys. Rev. B, 65, 035109, 2002.
21. Dreizler, R.M. and Gross, E.K.U., Density Functional Theory: An Approach to the
Quantum Many-Body Problem, Springer-Verlag, Berlin, 1990.
22. Briggs, E.L., Sullivan, D.J., and Bernholc, J., Real-space multigrid-based approach
to large-scale electronic structure calculations, Phys. Rev. B, 54, 14362, 1996.
23. Baroni, S., Dal Conso, A., De Gironcoli, S., and Giannozzi, P., Plane wave-based
calculations have been performed using the PWscf package,
http://www.pwscf.org.
24. Ceperley, D.M. and Alder, B.J., Ground state of the electron gas by a stochastic
method, Phys. Rev. Lett., 45, 566, 1980; Perdew, J.P. and Zunger, A., Self-interaction
correction to density-functional approximations for many-electron systems, Phys. Rev.
B, 23, 5048, 1981.
25. Fuchs, M. and Scheffler, M., Ab initio pseudopotentials for electronic structure cal-
culations of poly-atomic systems using density-functional theory, Comput. Phys.
Commun., 119, 67, 1998; Troullier, N. and Martins, J.L., Efficient pseudopotentials
for plane-wave calculations, Phys. Rev. B, 43, 1993, 1991.
26. Kleinman, L. and Bylander, D.M., Efficacious form for model pseudopotentials, Phys.
Rev. Lett., 48, 1425, 1982.
27. Zoroddu, A., Bernardini, F., Ruggerone, P., and Fiorentini, V., First principles predic-
tion of structure, energetics, formation enthalpy, elastic constants, polarization, and
piezoelectric constants of AlN, GaN, and InN: comparison of local and gradient-
corrected density-functional theory, Phys. Rev. B, 64, 045208, 2001.
28. Bernardini, F., Fiorentini, V., and Vanderbilt, D., Spontaneous polarization and
piezoelectric constants of III-V nitrides, Phys. Rev. B, 56, R10024, 1997; Accurate
calculation of polarization-related quantities in semiconductors, Phys. Rev. B, 63,
193201, 2001.
29. Sai, N. and Mele, E.J., Microscopic theory for nanotube piezoelectricity, Phys. Rev.
B, 68, 241405(R), 2003.
30. Nye, J.F., Physical Properties of Crystals, Claredon Press, Oxford, 1985., Chap. 7.
31. Hernandez, E., Goze, C., Bernier, P., and Rubio, A., Elastic properties of C and B
x
C
y
N
z
composite nanotubes, Phys. Rev. Lett., 80, 4502, 1998.
32. Resta, R., private communication; Kleinman, L., Comment on the average potential
of a Wigner solid, Phys. Rev. B, 24, 7412, 1981.
33. Posternak, M., Baldereschi, A., Catellani, A., and Resta, R., Ab initio study of the
spontaneous polarization of pyroelectric BeO, Phys. Rev. Lett., 64, 1777, 1990.
34. Baldereschi, A., Baroni, S., and Resta, R., Band offsets in lattice-matched hetero-
junctions: a model and first-principles calculations for GaAs/AlAs, Phys. Rev. Lett.,
61, 734, 1988.
35. Kane, C.L. and Mele, E.J., Broken symmetries in scanning tunneling images of carbon
nanotubes, Phys. Rev. B, 59, R12759, 1999.
36. Fowler, R.H. and Nordheim, L.W., Electron emission in intense electric fields, Proc.
Roy. Soc. Lond., A119, 173, 1928.
37. Han, S. and Ihm, J., Role of the localized states in field emission of carbon nanotubes,
Phys. Rev. B, 61, 9986, 2000.
38. Bengtsson, L., Dipole correction for surface supercell calculations, Phys. Rev. B, 59,
12301, 1999.
39. Golberg, D., Dorozhkin, P.S., Bando, Y., Dong, Z.-C., Tang, C.C., Uemura, Y., Grobert,
N., Reyes-Reyes, M., Terrones, H., and Terrones, M., Structure, transport and field-
emission properties of compound nanotubes: CN
x
vs. BNC
x
(x < 0.1), Appl. Phys. A,
76, 499, 2003.
40. Golberg, D., Dorozhkin, P.S., Bando, Y., Dong, Z.-C., Grobert, N., Reyes-Reyes,
M., Terrones, H., and Terrones, M., Cables of BN-insulated B-C-N nanotubes, Appl.
Phys. Lett., 82, 1275, 2003.
41. Golberg, D., Bando, Y., Kurashima, K., and Sato, T., Ropes of BN multi-walled
nanotubes, Solid State Commun., 116, 1, 2000; HRTEM and electron diffraction
studies of B/N-doped C and BN nanotubes, Diam. Relat. Mater., 10, 63, 2001; Lee,
R.S., Gavillet, J., Lamy de la Chapelle, M., Loiseau, A., Cochon, J.-L., Pigache,
D., Thibault, J., and Willaime, F., Catalyst-free synthesis of boron nitride single-wall
nanotubes with a preferred zig-zag configuration, Phys. Rev. B, 64, 121405(R), 2001.
22
Multiscale Modeling
of Stress Localization
and Fracture in
Nanocrystalline
Metallic Materials
Vesselin Yamakov, Dawn R. Phillips,
Erik Saether, and Edward H. Glaessgen
CONTENTS
22.1 Introduction
22.2 Configuration Model
22.3 Molecular Dynamics Model
22.4 Shear Strength of a Grain Boundary
22.5 FEM Simulation
Problems
Acknowledgments
References
22.1 INTRODUCTION
Uneven stress distribution and stress localization during deformation are key factors
for fracture and failure in polycrystalline and nanocrystalline metals. The inho-
mogeneous polycrystalline microstructure that consists of grains of different sizes
and shapes joined together at different angles and forming various types of grain
boundaries (GBs) creates inhomogeneous deformation fields under homogeneous load-
ing. There are a number of factors responsible for the appearance of inhomogeneous
deformation inside the polycrystal. The coexistence of grains of different sizes and
orientations with anisotropic elastic properties is one factor. The difference in the
structures and properties of the GBs between grains of various misorientations is a
second factor for uneven stress distribution. A third factor is grain-boundary sliding
(GBS), i.e., the rigid translation of one grain relative to another at the GB interface.
When grains deform, GBS is an inevitable process as a result of the relative move-
ments and rearrangements of the grains.
1
GBS is a strongly inhomogeneous mode
of deformation localized at a very narrow interface layer, thus creating very strong
shear forces. This, together with the fact that the shear strength of a general GB
between grains of high misorientation angle (high-angle GB) is much lower than
the shear strength of the perfect crystal,
2,3
causes GBS to be a prominent deformation
mode. Particularly in nanocrystalline metals, because of their extremely small grain
size (less than 100 nm) resulting in a dense network of GBs, GBS assisted through
stress-driven GB diffusion for grains smaller than 20 nm
46
and through dislocation
slip in larger grains
79
was found to be a major mode of deformation. During GBS,
the load transfer between the sliding surfaces is significantly reduced, and the load
is redirected to other places more resistant to sliding such as the triple junctions
where three GBs meet and GBS cannot be accommodated. In this way, sliding creates
redistribution of the load and the appearance of stress localization in the microstruc-
ture that, in the absence of an efficient accommodation mechanism, can lead to void
formation and microcracking starting at the triple junctions.
10,11
This chapter describes a multiscale modeling strategy that is used to study the
effect of GBS for stress localization in a polycrystalline microstructure. Although
they reveal system behavior at atomic-level resolution, molecular dynamics (MD)
calculations of large systems of grains are computationally prohibitive to perform,
and a more efficient analysis technique is sought. Finite element method (FEM)
analysis is an obvious choice, but the FEM models cannot simulate the deformation
mechanisms found within the systems of grains a priori. Thus, the FEM models
must be tuned in order to reproduce the stress localization observed in the atomistic
simulations. For this purpose, a nanocrystalline model of bimodal grain-size distri-
bution was constructed and is convenient for study by both MD and FEM simulations.
The MD simulations revealed the behavior of this model with atomic-scale details.
Additional MD simulations on a bicrystal model were performed to extract the elastic
and yield properties of the GBs presented in the model. These GB properties were
then used to tune the FEM model to reproduce closely the behavior of the nano-
crystalline model to match with the MD simulation results. The FEM model then
served as part of a multiscale modeling strategy to extrapolate the MD-derived
information to larger scales to study the failure properties of nanocrystalline metals.
In this chapter, details of the configuration model used in both MD and FEM
simulations are discussed first. The specific features of the MD model are then presented.
Next the FEM version of the model is discussed, and the results of the FEM simulation
are compared with those from the MD simulation. Finally a large-scale FEM simulation
of a typical microstructure of 100 grains is analyzed to reveal the stress distribution due
to grain boundary sliding in a more general polycrystalline system.
22.2 CONFIGURATION MODEL
To investigate the effect of GBS on stress distribution in a polycrystal of inhomo-
geneous grain size, a configuration model, presented in Fig. 22.1(a) and (b), was
used. This configuration was used in both MD and FEM simulations with the
FIGURE 22.1 (a) Snapshot of the thermally equilibrated polycrystalline microstructure
designed to study the effect of grain boundary sliding on stress distribution in a polycrystal of
bimodal grain size by molecular dynamics simulation. The grain boundaries are shown as
darker lines of atoms that were identified to be in a disordered (noncrystalline) local environ-
ment. The dashed lines mark the actual system box dimensions, while for better visualization,
the microstructure is slightly extended using periodic boundary conditions. (b) Repeating unit
used in finite element simulations to model grain structure shown in (a). As there is no
characteristic length scale in the finite element model, the dimensions in (b) are relative,
characterized by a: b 1:4 and consistent with (a).
[110]
[110]
[110]
[110]
[001]
[001]
[001]
[001]
(a)
20 nm
8
0

n
m
X
Y
(
7
,
7
,
1
0
)
9
9

S
T
G
B
(
5
5
7
)
9
9

S
T
G
B
45
XX
= 1 GPa
u
y
= const
u
x
= 0
u
y
= 0
Y
X
e
x
= 1.23%
a
b
(b)
additional goal of tuning the parameters of FEM to reproduce as closely as possible
the MD simulation. The model is a periodically repeating unit of a four-grain
octagonsquare configuration that, when replicated in the x- and y-directions, rep-
resents a polycrystal of a bimodal grain-size distribution. The advantages of this
configuration are the following: (1) The model allows exploration of the effects of
large variations in the ratio a : b between the grain sizes (Fig. 22.1(b)), which for this
study was set to 1:4; (2) the model is both simple enough to allow for MD simulation
and informative enough to be used in the FEM simulation to compare with the MD
results; and (3) the configuration has a four-fold symmetry (against 90q rotation),
which further simplifies the analysis. In both MD and FEM simulations, this con-
figuration is deformed by applying stress or strain, respectively, along the x-direction.
The deformation conditions are chosen such that the grains deform elastically, while
the GBs deform plastically by GBS. An interatomic potential representative of
aluminum (Al) is used in the MD model, and an isotropic continuum is assumed
for the grains in the FE model to isolate the effect of crystal anisotropy on the stress
distributions. Though this would result in some inevitable discrepancy in the results,
it helped to identify the amount of stress distribution due to GBS alone. Fixed
boundary conditions shown in Fig. 22.1(b) that for an isotropic media mimic the
periodic boundary conditions in the MD simulation shown in Fig. 22.1(a) were
applied in the FE simulation.
22.3 MOLECULAR DYNAMICS MODEL
To investigate the effect of GBS on the stress distribution in a polycrystal of inho-
mogeneous grain size, atomistic MD simulations were used where the only predefined
input quantity was the interactive potential between individual atoms, which, follow-
ing Newtonian dynamics, defines their motion. The embedded atom method (EAM)
many-body potential
12
was fitted to reproduce closely the elastic and thermodynamic
properties of a perfect Al crystal and used in the simulation. A textured or columnar
microstructure model with periodic boundary conditions in all three dimensions to
mimic bulk conditions
13
was selected. While providing a fully three-dimensional (3-
D) treatment of the underlying physics, this model makes it possible to simulate
relatively large grains, because only a few lattice planes need to be considered in the
periodically repeated texture direction. The texture z-axis is along a [110] crystallo-
graphic orientation, which enables realistic dislocation slip dynamics along six avail-
able slip systems in two [111] slip planes in each grain.
14
These slip systems are
adequate to accommodate any two-dimensional deformation in the x-y plane of the
simulated structure. The grain boundaries, formed between the grains in this model,
are [110] tilt GBs that have been studied extensively in the past 15 years, both
experimentally
1518
and by simulations.
1721
In spite of some drawbacks inherent to
this model when compared to a fully 3-D structure, as discussed in Reference 7, the
model was remarkably successful in predicting the highly unexpected deformation
twinning in nanocrystalline Al
22,23
and the transition from partial to perfect slip in
nanocrystalline fcc metals.
14,22
Both of these phenomena have since gained solid
experimental support.
2426
A four-grain octagon-square configuration was used as a test model. A snapshot
of the four-grain octagon-square microstructure after thermal equilibration at 100
K is given in Fig. 22.1(a). The configuration represents a periodically repeating
unit of large octagonal grains of size b 80 nm encapsulating small square grains
of size a 20 nm. The thickness of the system in the texture direction was set equal
to 5 (1, 1, 0) atomic planes (1.43 nm) as in the previous columnar models.
13,14,22,23
Within these dimensions, the system contains 1.7 million atoms. By rotating one
of the two octagonal grains at 90q relative to the other (the crystallographic orien-
tations are shown in Fig. 22.1), the four 45q inclined diagonal GBs become 90q
699 symmetric-tilt GBs (STGBs) for which the atomic structure in Al is known
from the literature
18
. The two types of 699 STGBs with (5, 5, 7) and (7, 7, 10) GB
planes (see Fig. 22.1(a)) are structurally very similar and practically undistinguish-
able from the point of view of their mechanical properties, which preserves the
four-fold symmetry (against 90q rotation) of the structure. In addition, the two small
square grains are rotated by 45q relative to the large grains, thus forming high-
angle asymmetric tilt GBs
15,16
(Fig. 22.1(a)).
To initiate GB sliding, the microstructure was loaded with a uniform tensile
stress of 1 GPa along the x-axis in Fig. 22.1(a). The stress was applied using the
ParrinelloRahman
27
constant stress technique combined with a NoseHoover
thermostat
28
for constant-temperature simulation. This stress is well below the
threshold stress of about 2 GPa
7,14
needed to start nucleation of dislocations from
the GBs, which are the only possible dislocation sources in grains of nanometer
size.
7,22
Thus the dislocation activity, as a possible accommodation mechanism,
was readily suppressed. By running the simulations at a very low temperature of
100 K (a value still high enough to avoid quantum mechanical effects existing in a
real structure that cannot be captured by the classical MD technique), GB diffusion,
as another possible accommodation mechanism was also eliminated. Low temper-
ature also prevents grain growth, a process that otherwise would be very strong in
such a system of a bimodal grain-size distribution. Restricting the dislocation activity
and GB diffusion helps to map the MD model nearer to our FEM simulations, which,
at the present stage, do not include these mechanisms.
During loading, the system reached an equilibrium strain of 1.23%. The strain
was elastic within the grains and plastic at the GBs, thus initiating GB sliding. At
this stage, to present the stress distribution in the system, two-dimensional stress
maps were created. The two-dimensional resolution of the maps is an area of 6 u 6
lattice parameters (2.43 u 2.43 u 1.43 nm
3
volume enclosing 432 atoms) over which
the local stress was averaged. The stress maps were stored every 1 ps over a period
of 22 ps after reaching equilibrium (approx. 40,000 MD steps) and then averaged
in time to smooth out the fluctuations always present in a system of such small size.
These averaging procedures in space and time gave a good estimate of the local
stress calculated by the virial expansion technique in ParrinelloRahman stress
calculation.
27
All the stress maps for the three stress components in the x-y plane of
deformation (i.e., normal to the texture z-axis), V
xx
, V
yy
, and V
xy
, showed distinctive
stress distribution and stress localization that were then compared with an FEM
simulation of the same configuration, as will be discussed.
22.4 SHEAR STRENGTH OF A GRAIN BOUNDARY
Knowledge of the shear strength of the GBs presented in the system is crucial for
understanding the stress distribution due to GBS. This shear strength is also required
as an input for the FEM model discussed later. For this purpose, a separate MD
simulation of a bicrystal cubic system, presented in Fig. 22.2, was used. The same
interatomic potential
12
as in the octagonsquare model was applied.
The two crystals were crystallographically oriented in such a way as to form a
(7, 7, 10) 699 STGB (Fig. 22.2). This type of GB, together with the very similar
(5, 5, 7) 699 STGB, was expected to exhibit most of the GBS in the octagonsquare
model (Fig. 22.1) due to the maximum resolved shear stress on the planes inclined
at 45q to the tensile direction. In the simulation, the system box with periodic
boundary conditions was allowed to shear in all three directions in addition to
expansion or contraction, making it possible to estimate the shear strain directly.
The system was loaded with a symmetric shear stress (V
xy
V
yx
) to prevent torque
forces and to initiate GBS in the direction consistent with the sliding direction
between the octagons in the model. In addition, a control simulation was performed
on a perfect crystal of the same size and orientation ([5, 5, 7]/[7, 7, 10]/[1, 1, 0]) to
obtain the shear strain of a perfect crystal under the same simulation conditions.
The perfect crystal strain was then subtracted from the bicrystal strain, and the
resulting strain due to the GB alone is presented in Fig. 22.3 as a function of load
for three different temperatures.
Following Schiotz et al.,
4
the yield stress was defined as the stress where the
strain starts to depart from linearity, a definition convenient for use in MD simula-
tions. For the temperature of T 100 K used in the simulation of the octagonsquare
model the obtained yield stress is V
y
0.2 GPa. As expected, the yield stress
decreases with increasing temperature of the system, becoming 0.14 GPa at 200 K
FIGURE 22.2 Bicrystal cubic microstructure used to determine the shear strength of 699
symmetric tilt grain boundary formed in the octagonsquare system.
[7,7,10]
[7,7,10]
3[5,5,7]
20[1,1,0]
12.09 nm
11.46 nm
11.40 nm
GB
Y
Z
X
and 0.12 GPa at 300 K. For comparison, the theoretical shear strength of a perfect Al
crystal was estimated at around 2.8 GPa.
29
Thus, the GBs are shown to be more than
an order of magnitude weaker to shear than the perfect crystal grains. Applying a
tensile load of 1 GPa along the x-axis in the octagonsquare model (Fig. 22.1(a))
resulted in a resolved shear stress of 0.5 GPa along the diagonal 699 GB planes. This
stress is more than twice as large as the GB yield stress, as defined here (see Fig. 22.3),
but still well below the shear strength of the perfect crystal. This ensures a regime
of deformation where the GBs would experience plastic sliding while the crystal
grains would deform elastically.
22.5 FEM SIMULATION
For the finite element simulations, a modified version of the FRANC code developed
at Cornell University
30
was used and specifically designed to study crack initiation
and propagation in metallic polycrystals with explicit representations of grains and
grain boundaries. Each of the grains in the model was given isotropic, elastic material
properties. The properties used were Youngs modulus, E 74.8 GPa, and Poissons
ratio, Q 0.346. These values were calculated averages
31
from the anisotropic elastic
constants of the interatomic potential
12
used in the MD approach. To accurately
account for the entire structure, a repeating unit, similar in geometry to the one used
in the MD simulation, was chosen (Fig. 22.1(b)). As in the MD model, the ratio of
the lengths of the grain boundaries was a : b 1: 4. Periodic boundary conditions
are necessary to accurately compare results to those obtained from MD simulations.
The boundary and loading conditions shown in Fig. 22.1(b) are consistent with
periodic boundary conditions for an isotropic material.
FIGURE 22.3 Grain boundary shear strain versus stress extracted from the bicrystal model.
The straight lines tangential to the curves at their origin mark the linear elastic regime. The
arrows show the yield stress for each of the three simulated temperatures defined as the stress
where the strain starts to depart from linearity.
T = 300K
T = 200K
T = 100K
0.8
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4
xy
(GPa)
(
c
r
)
x
y

(
%
)
In this work, cohesive zone models (CZMs) were used to characterize grain
boundary behavior.
32
Cohesive zone models assume cohesive interactions of the
material around a grain boundary and permit the appearance of fracture surfaces in
a continuum.
33
To simulate the sliding behavior along grain boundaries, two inde-
pendent cohesive zone models were chosen for the normal and shear components
of the traction and displacement. To permit sliding, a perfectly plastic relationship
was chosen for the shear model with yield stress of 0.2 GPa; this stress is consistent
with the numbers obtained from the MD criterion discussed previously. To restrict
opening, the normal CZM was specified as having linear elastic behavior with high
stiffness. The analysis was performed for 1.23% applied strain. This applied strain
corresponds to the 1 GPa applied stress in the MD simulations.
Applying a uniform tensile stress of 1 GPa creates resolved shear stresses larger than
the GB yield stress but lower than the yield stress of the grains, producing plastic
sliding at the GBs and elastic deformation in the grain interiors. GBS reduces the load
transfer from one grain to another and redistributes the stress. The V
xx
, V
yy
, and V
xy
stress distributions for the octagon-square model are shown in Fig. 22.4(af ). The
results from the FE simulations presented in Fig. 22.4(ac) are compared with the
stress maps obtained from the MD simulation and presented in Fig. 22.4(df ). Both
models show that the small square grains take most of the load produced from the
tensile stress (Fig. 22.4(a) and (d)) and from the Poisson contraction perpendicular to
the tensile x-direction (Fig. 22.4(b) and (e)). This contraction creates a strong tensile
stress at the sides of the square grains normal to the tensile direction (regions marked
(1) in Fig. 22.4(b) and (e)). The close similarity between the MD and FEM stress
distribution and the fact that the FEM model was isotropic indicate that this stress
distribution is solely due to GBS. Through GBS, the grains become decoupled in their
elastic deformation. At the GBs, where only two grains meet, this decoupling creates
sliding, but at the triple junctions, where three grains meet, strong incompatibility
stresses are produced. This is seen in both FEM and MD results (Fig. 22.4(c) and (f )).
The most direct evidence for the presence of GBS is seen on the MD V
xy
stress
map in Fig. 22.4(f ). The four diagonal GBs showed strong shear stress in the two
opposite directions (with stress values presented in red and blue). Figure 22.5 is a
schematic representation of the grain sliding responsible for the observed shear stress
distribution. Grains (1) and (3) are displacing toward each other, compressing the
small grain (5) in the y-direction, while grains (2) and (4) are displacing apart and
stretching the grain (5) in the x-direction. The displacement of the large grains results
in sliding at the GBs, schematically shown in Fig. 22.5 as couples of oppositely
directed arrows on both sides of the diagonal GBs. The direction of each arrow
corresponds to the local displacement of each grain at the GB. This displacement
produces strong local shear stress at the diagonal GBs, as seen in Fig. 22.4(c) and (f ).
In spite of the overall good match between the stress distribution from MD and
FEM simulations, there are also significant mismatches that can be discovered after
a careful examination of Fig. 22.4(af ). For example, in the MD simulation, there
is no full symmetry of the stresses produced on the two small grains, and there are
some deviations in the stresses on the large grains as well (most pronounced in
Fig. 22.4(d)). Possibly the main reason for these deviations is the small but noticeable
anisotropy in the MD model that results from the interatomic potential. This potential
was fitted
12
to reproduce closely the elastic constants including the anisotropy of
monocrystalline Al. An anisotropic FEM simulation with the proper periodic bound-
ary conditions is needed to achieve a better comparison.
Another source of mismatch between the two models is the inevitable existence
of GB dislocations in the MD model. In a polycrystalline MD model, absolutely perfect
GBs are impossible to sustain during loading, even if they were produced initially.
FIGURE 22.4 Stress contours from the finite element model (ac) and the corresponding
stress maps from the molecular dynamics model (df ) for V
xx
, V
yy
, and V
xy
stress components.
Positive and negative stresses are indicated relative to the average background stress. In (a),
(b), (d), and (e), positive stress is defined as the stress in compression and negative stress as
the stress in tension compared to the background stress. In (c) and (f ), positive and negative
shear corresponds to shear directions at the diagonal grain boundaries. The stress localization
of V
yy
at the grain boundaries of the small grain marked as (1) in (b) and (e) are significant.
Shear stress concentrations created at the triple junctions are marked (2) in (c) and (f ). A few
grain boundary dislocations that appeared in the molecular dynamics simulation and which
create disturbances in the grain boundary stress field are indicated by (3) in (f ).
XX

YY

XY
2.3 0.7 0.6 1.2 0.1 1.7 0.4 0.0 0.4
2.4 1.1 0.3 1.2 0.1 1.7 0.4 0.0 0.4
GPa GPa GPa
GPa GPa GPa
(a) (b) (c)
(d) (e) (f)
(1)
(1)
(3)
(3)
(2)
(2)
FEM
MD
The same is true for real materials such as nanocrystalline metals, in which GB
dislocations are known to contribute to the deformation.
10
However, at this stage, GB
dislocations are not considered in the FEM model. The GB dislocations present in
the MD model have long-range elastic fields that can significantly modify the stress
distribution in close proximity to the GBs, thus altering it from the FEM result. Such
an effect produced by GB dislocations is marked as (3) in Fig. 22.4(f ). Also, the
phonon waves always existing in finite temperature MD simulation create periodic
modulations in the MD stress maps (Fig. 22.4(df )) that are nonexistent in the
FEM stress distributions (Fig. 22.4(ac)). Even averaging the stress over a 20-ps
time interval was not enough to smooth these fluctuations. One reason may be that
the periodic boundary conditions imposed on the system, together with its high con-
figuration symmetry, create stationary waves that cannot be smoothed out over such a
short period of time. All these factors make the MD maps shown in Fig. 22.4 much
richer and more complicated compared to the substantially more idealized FEM results.
However, the generally close quantitative match between the stress distribution
obtained from MD and FEM simulations of the octagonsquare configuration
(Fig. 22.4) gives confidence that implementing the properly parameterized CZM
elements in the FEM model correctly reproduces GBS for a polycrystalline specimen.
With this confidence, FEM can now be used to simulate a typical nanocrystalline
microstructure. For this purpose, a sample with 100 randomly shaped and sized grains
represented by Voronoi polygons
34
was constructed. The grain configuration, along
with the boundary and loading conditions, is shown in Fig. 22.6.
For 1.23% applied strain, the V
xx
stress distribution is shown in Fig. 22.7. The
similarities in grain boundary behavior between the randomly configured model
and the idealized octagonsquare configuration are highlighted. This larger scale
FIGURE 22.5 Schematic representation of the sliding directions of the four octagonal grains
(bold arrows) and the inserted stresses on the small square grain in the middle (dashed arrows).
The local displacements of the material at the grain boundaries producing grain boundary
sliding are presented by couples of opposite pointing arrows on both sides of the boundaries.
XX

XX
Y
X
(3)
(2)
(1)
(5)
(4)
FIGURE 22.6 Material model with 100 randomly shaped and sized grains represented by
Voronoi polygons.
FIGURE 22.7 (Color figure follows p. 12.) Stress contour V
xx
for 100-grain material subjected
to 1.23% strain with similarities to the idealized model highlighted.
u
x
= 0
u
y
= const
u
y
= 0 e
x
= 1.23%
Y
X
XX
2.3 0.7 0.6
GPa
simulation shows that stress localizations due to GBS are common within the
nanocrystals, and local stress can exceed several times the background stress, thus
marking spots for a possible crack nucleation. It is not surprising that these stress
concentrations are mostly localized at the triple junctions where GB sliding cannot
be accommodated.
35
Summarizing the results from this section, it was shown that simulating the
behavior of geometrically identical nanocrystalline structures by MD and FEM
techniques can serve two purposes. First, it helps to tune the parameters of the larger
scale FEM simulation to the results from the lower scale MD simulation, thus
incorporating the effects of atomistic mechanisms, present in MD and essential for
the deformation at nanoscales, into a continuum model. Second, such a parallel use
of two very different techniques can help to identify more unambiguously the
underlying mechanism behind a certain phenomenon. Reproducing a stress distri-
bution by isotropic FE simulation similar to the distribution obtained from MD
simulation where anisotropy is inherent from the interatomic potential and cannot
be turned off proves that GBS played the major role in this case. As a result, even
in a material of weak anisotropy, such as Al, substantial stress localization can be
produced due to GBS. It should also be noted that the effects of anisotropy and the
intragranular dislocation slip are much more limited in the deformation of nanoc-
rystalline materials compared to coarse-grained materials. Thus, neglecting both of
these effects in the nanocrystalline FEM model presented here appears to be rea-
sonable.
A combined multiscale implementation of two modeling techniques, MD and FEM,
has been demonstrated at different length-scale resolutions on two geometrically
equivalent nanocrystalline systems. The results, after comparing these two simula-
tions, are very instructive. First, a method was successfully applied to match the
parameters in the constitutive equations (in this case the CZM) used in the FEM
model with quantities extracted from the MD simulation (such as GB yield stress).
By this method, the FEM was tuned to the MD simulation. Second, in an MD
simulation where the resolution is at the level of individual atoms, it is usually very
difficult to identify unambiguously the underlying mechanism responsible for a given
process. By reproducing the MD simulation experiment by a continuum FEM model,
where the mechanisms are implemented through constitutive equations and can be
easily controlled, the underlying mechanism can be revealed with higher certainty.
For example, in the octagonsquare model presented here, the stress distribution
obtained by MD simulation cannot be attributed readily to GBS, as there is also
anisotropy that can produce analogous effects. After reproducing the similar simu-
lation experiment by FEM using isotropic material parameters, but allowing for GBS
with the equivalent sliding resistance as in the MD model and comparing the results,
it becomes possible to identify what part of the stress distribution is due to GBS
alone and what part is due to other sources such as anisotropy and GB dislocations.
It has been also demonstrated that now, because of the advanced computational
technology, high-end MD simulations can overlap well with FEM simulations on
small but still informative systems to allow direct comparison of the results. This
possibility suggests a new way of bridging length and time scales and is an alternative
to the direct multiscale models, where the simulated system is divided into regions
of different length and time scales that are treated with different simulation tech-
niques simultaneously.
36
Instead of trying to combine the two inherently very dif-
ferent simulation techniques into one multiscale modeling program, it is now pos-
sible to use the overlapping region to tune the larger scale model to the lower scale
one by direct comparison. This approach has the advantage that it preserves the
uniqueness and the integrity of each of the models and avoids creating overcompli-
cated software codes with very restricted applicability. This approach also avoids
potential artifacts of the interface between regions of different length and time scales
existing in the direct multiscale models, such as unrealistic phonon reflection or
dislocation repulsion from the interface, etc.
36
Thus, it becomes feasible to model
much larger polycrystalline systems than could be handled by MD alone.
Finally, the large-scale implementation of the MD-tuned FEM simulation on a
typical 100-grain nanocrystalline microstructure showed that GBS, by redistributing
the load in the system, can produce stress localization several times higher than the
background stress under external uniform load. These stress localizations, under
suitable conditions, may become nuclei for microcracking and void formation. These
results may help to explain the generally observed much higher brittleness of nanoc-
rystalline metals compared to the coarse-grained metals.
37
PROBLEMS
Selected problems pertaining to the multiscale combination of discrete MD and
continuum FEM analyses are presented below. It is assumed that the reader has a
formal knowledge or practical familiarity with either molecular dynamics or finite
element methods. Fundamental issues involving the underlying physics or details
of the computational methods have been, by necessity, not elaborated in the present
chapter but are assumed available to the reader through the listed references or
through general literature searches. Solutions to the following problems are aimed
at providing the reader with a good understanding of many of the important
theoretical issues involved in MD-FEM multiscale analysis methods.
1. Identify the differences between sequential and simultaneous multiscale
methods. What are the advantages and disadvantages of each?
2. Discuss the considerations related to determining the smallest length scale
for which a continuum model is applicable. Under what conditions do
continuum mechanics assumptions break down? List the sources of error
associated with the application of continuum mechanics constructs at very
small length scales.
3. List several nonphysical artifacts that can be generated in modeling small
molecular dynamics systems.
4. Discuss several deformation mechanisms that are explicitly modeled by
molecular dynamics analyses that can only be considered in an average
sense in finite element analyses.
5. Why is a columnar microstructural configuration needed in molecular
dynamics analyses to properly correlate with a two-dimensional finite
element analysis? In general, what are some of the advantages and dis-
advantages of using a two-dimensional model in these solution methods?
6. Discuss the qualitative relationship between the size of a grain and its
dominant deformation mechanism. Under what conditions is grain boundary
sliding the dominant deformation mechanism of a grain? What deformation
mechanisms are suppressed in small grains? Hint: perform a literature search
on the HallPetch relationship to gain insight to this question.
ACKNOWLEDGMENTS
The work of V. Yamakov and D.R. Phillips was performed at NASAs Langley
Research Center and sponsored through cooperative agreement NCC-1-02043 with
the National Institute of Aerospace and contract NAS1-00135 with Lockheed Martin
Space Operations.
REFERENCES
1. Langdon, T.G., Identifying creep mechanisms at low stresses, Mater. Sci. Eng. A,
283, 266273, 2000.
2. Wolf, D. and Jaszczak, J.A., Tailored elastic behavior of multilayers through con-
trolled interface structure, J. Computer-Aided Mater. Design, 1, 111148, 1993.
3. Adams, J.B., Wolfer, W.G., and Foiles, S.M., Elastic properties of grain boundaries
in Cu and their relationship to bulk elastic constants, Phys. Rev. B, 40, 94799484,
1989.
4. Schiotz, J., DiTolla, F.D., and Jacobsen, K.W., Softening of nanocrystalline metals
at very small grain sizes, Nature, 391, 561563, 1998.
5. Van Swygenhoven, H., Polycrystalline materialsgrain boundaries and dislocations,
Science, 296, 6667, 2002.
6. Yamakov, V., Wolf, D., Phillpot, S.R., and Gleiter, H., Grain-boundary diffusion creep
in nanocrystalline palladium by molecular-dynamics simulation, Acta Mater., 50,
6173, 2002.
7. Yamakov, V., Wolf, D., Phillpot, S.R., Mukherjee, A.K., and Gleiter, H., Crossover
in HallPetch behavior in nanocrystalline materials by molecular-dynamics simula-
tion, Phil. Mag. Lett., 83, 385393, 2003.
8. Yamakov, V., Wolf, D., Phillpot, S.R., Mukherjee, A.K., and Gleiter, H., Deformation-
mechanism map for nanocrystalline metals by molecular-dynamics simulation, Nature
Mater., 3, 4347, 2003.
9. Schiotz, J. and Jacobsen, K.W., A maximum in the strength of nanocrystalline copper,
Science, 301, 13571359, 2003.
10. Gutkin, M.Y., Ovidko, I.A., and Skiba, N.V., Strengthening and softening mecha-
nisms in nanocrystalline materials under superplastic deformation, Acta Mater, 52,
17111720, 2004.
11. Kumar, K.S., Suresh, S., Chisholm, M.F., Horton, J.A., and Wang, P., Deformation
of electrodeposited nanocrystalline nickel, Acta Mater, 51, 387405, 2003.
12. Mishin, Y., Farkas, D., Mehl, M.J., and Papaconstantopoulos, D.A., Interatomic
potentials for monoatomic metals from experimental data and ab initio calculations,
Phys. Rev. B, 59, 33933407, 1999.
13. Yamakov, V., Wolf, D., Salazar, M., Phillpot, S.R., and Gleiter, H., Length-scale
effects in the nucleation of extended lattice dislocations in nanocrystalline Al by
molecular-dynamics simulation, Acta Mater., 49, 27132722, 2001.
14. Yamakov, V., Wolf, D., Phillpot, S.R., and Gleiter, H., Dislocationdislocation and
dislocationtwin reactions in nanocrystalline Al by molecular-dynamics simulation,
Acta Mater., 51, 41354147, 2003.
15. Merkle, K.L., High-resolution electron microscopy of interfaces in FCC materials,
Ultramicroscopy, 37, 130152, 1991.
16. Merkle, K.L., Atomic structure of grain boundaries, J. Phys. Chem. Solids, 55, 991
1005, 1994.
17. Medlin, D.L., Foiles, S.M., and Cohen, D., A dislocation-based description of grain
boundary dissociation: application to a 90q 110 tilt boundary in gold, Acta Mater.,
49, 36893697, 2001.
18. Dahmen, U., Hetherington, J.D., OKeefe, M.A., Westmacott, K.H., Mills, M.J., Daw,
M.S., and Vitek, V., Atomic structure of a 699 grain boundary in Al: a comparison
between atomic-resolution observation and pair-potential and embedded-atom simu-
lations, Phil. Mag. Lett., 62, 327335, 1990.
19. Wolf, D., Structureenergy correlation for grain boundaries in F.C.C. metals III.
Symmetrical tilt boundaries, Acta Metal., 38, 781790, 1990.
20. Rittner, J.D. and Seidman, D.N., 110 Symmetric tilt grain-boundary structures in
FCC metals with low stacking-fault energies, Phys. Rev. B, 54, 69997015, 1996.
21. Nishitani, S.R., Ohgushi, S., Inoue, Y., and Adachi, H., Grain boundary energies of
Al simulated by environment-dependent embedded atom method, Mat. Sci. Eng. A,
309310, 490494, 2001.
22. Yamakov, V., Wolf, D., Phillpot, S.R., Mukherjee, A.K, and Gleiter, H., Dislocation
processes in the deformation of nanocrystalline Al by molecular-dynamics simulation,
Nature Mater., 1, 4548, 2002.
23. Yamakov, V., Wolf, D., Phillpot, S.R., and Gleiter, H., Deformation twinning in
nanocrystalline Al by molecular-dynamics simulation, Acta Mater., 50, 50055020,
2002.
24. Chen, M., Ma, E., Hemker, K.J., Sheng, H., Wang, Y., and Cheng, X., Deformation
twinning in nanocrystalline Al, Science, 300, 12751277, 2003.
25. Liao, X.Z., Zhou, F., Lavernia, E.J, Srinivasan, S.G., Baskes, M.I., He, D.W., and
Zhu, Y.T., Deformation mechanism in nanocrystalline Al: partial dislocation slip,
Appl. Phys. Lett., 83, 632634, 2003.
26. Liao, X.Z., Zhou, F., Lavernia, E.J., He, D.W., and Zhu, Y.T., Deformation twins in
nanocrystalline Al, Appl. Phys. Lett., 83, 50625064, 2003.
27. Parrinello, M. and Rahman, A., Polymorphic transitions in single crystals: a new
molecular dynamics method, J. Appl. Phys., 52, 71827190, 1981.
28. Nose, S., A unified formulation of the constant temperature molecular dynamics
method, J. Chem. Phys., 81, 511519, 1984.
29. Ogata, S., Li, J., and Yip, S., Ideal pure shear strength of aluminum and copper,
Science, 298, 807811, 2002.
30. Iesulauro, E., Decohesion of grain boundaries in statistical representations of alumi-
num polycrystals, Cornell University Report, 0201, 2002.
31. Hirth, J.P. and Lothe, J., Theory of Dislocations, 2nd ed., John Wiley & Sons, New
York, 1992.
32. Tvergaard, V. and Hutchinson, J.W., The relation between crack growth resistance
and fracture process parameters in elastic-plastic solids, J. Mechan. Phys. Solids, 40,
13771397, 1992.
33. Klein, P. and Gao, H., Crack nucleation and growth as strain localization in a virtual-
bond continuum, Engin. Fracture Mechan., 61, 2148, 1998.
34. ORourke, J., Computational Geometry in C, 2nd ed., Cambridge University Press,
2001.
35. Shobu, K., Tani, E., and Watanabe, T., Geometrical study on the inhomogeneous
deformation of polycrystals due to grain-boundary sliding, Phil. Mag. A, 74, 957964,
1996.
36. Tadmor, E.B., Phillips, R., and Ortiz, M., Mixed atomistic and continuum models of
deformation in solids, Langmuir, 12, 45294534, 1996.
37. Wang, Y., Chen, M., Zhou, F., and Ma, E., High tensile ductility in a nanostructured
metal, Nature, 419, 912915, 2002.
23
Modeling of Carbon
Nanotube/Polymer
Composites
Gregory M. Odegard
CONTENTS
23.1 Introduction
23.2 Carbon Nanotube/Polymer Interface
23.3 Micromechanics
23.3.1 Analytical Micromechanical Models
23.3.2 Numerical Micromechanical Models
23.4 Molecular Models
23.4.1 Ab Initio Simulations
23.4.2 Molecular Dynamics
23.5 Example: SWNT/Polyimide Composite
23.5.1 Modeling Procedure
23.5.2 Results and Discussion
23.6 Example: SWNT/Polyethylene Composite
23.6.1 MD Simulations
23.6.2 StressStrain Curves from Simulation
23.6.3 Results
Problems
References
23.1 INTRODUCTION
The research and development of carbon fiber-reinforced polymer composites started
in the 1960s. Carbon fibers consist of tightly packed graphene sheets that are aligned
parallel to the fiber axis. The principal advantages of using carbon fibers include their
high modulus (up to 400 GPa), high strength (up to 3.4 GPa), low density (specific
gravity around 2.0), and low cost compared to many other reinforcement materials.
1
In 1985, Smalley and coworkers discovered the fullerene structure.
2
Fullerenes
are cage-like structures of carbon atoms that are composed of hexagonal and pentag-
onal faces. In 1991, Iijima
3
discovered carbon nanotubes. Nanotubes are elongated
fullerenes with walls consisting of graphene carbon structures (Fig. 23.1). Nanotubes
that consist of a single, closed graphene sheet are referred to as single-wall carbon
nanotubes (SWNTs). Nanotubes with multiple layers of closed graphene sheets are
referred to as multiwalled carbon nanotubes (MWNTs). The atomic structure of a
carbon nanotube is commonly described in terms of the chirality of the nanotube,
which is defined by the chiral vector D, which joins two equivalent atoms of a
graphene sheet. The vector is expressed as
(23.1)
where a
1
and a
2
are the unit vectors indicated in Fig. 23.2, the integers n and m
describe the number of steps along the vectors a
1
and a
2
, respectively, and the
boldface indicates a vector quantity. Nanotubes have extraordinary mechanical,
electrical, and thermal properties.
4,5
The composite materials community is showing tremendous interest in carbon
nanotubes for some of the same reasons carbon fibers generated interest 40 years
ago. That is, with a large aspect ratio (up to 3000), extraordinary properties (tensile
modulus and strength of up to 1 TPa and 200 GPa, respectively), and a relatively
low density (less than carbon fiber density), nanotubes could possibly serve as
ideal reinforcements for polymer resins. In addition, for a given volume fraction
of reinforcement, a composite reinforced with nanotubes contains a larger surface-
to-volume ratio of reinforcement than a composite reinforced with carbon fibers.
Therefore, more surface area for load transfer between the polymer matrix and
reinforcement exists for carbon nanotubes, which should, in theory, increase the
reinforcement efficiency.
Because of the high cost of carbon nanotubes (up to $500 per gram for purified
SWNTs), their research and development is quite expensive. In addition, trial-and-
error approaches are typically used to optimize the fabrication parameters for these
materials in the laboratory, with alternating steps of synthesis and characterization.
FIGURE 23.1 End section of SWNT.
D + n m a a
1 2
Modeling can be employed to facilitate the development of carbon nanotube mate-
rials. Incorporation of modeling leads to lower cost and quicker analysis of a wide
range of carbon nanotube/polymer composite materials compared to a purely exper-
imental approach. Two broad classes of modeling tools may be employed: micro-
mechanics and molecular modeling. While the continuum-based micromechanical
methods were originally developed for use on traditional, larger scale composite
materials, molecular modeling tools were developed for studying various forms of
molecular systems.
The objective of this chapter is to describe modeling methods for carbon nano-
tube/polymer composites. First some of the fundamental characteristics and recent
advances of carbon nanotube/polymer interfaces are reviewed, followed by a descrip-
tion of the four principal modeling tools for carbon nanotube/polymer composites:
analytical micromechanics, numerical micromechanics, ab initio simulations, and
molecular dynamics (MD) simulations. Finally examples of modeling techniques
are presented for SWNT/polyimide and SWNT/polyethylene composites.
23.2 CARBON NANOTUBE/POLYMER INTERFACE
The ability to transfer mechanical load between a carbon nanotube and the adjacent
polymer molecules in a composite material has been a matter of significant discus-
sion in the literature.
619
In general, two basic approaches have been proposed to
improve the strength of the interaction. The first approach involves forming a strong,
nonbonded interaction between the polymer and the nanotube without modifying
the nanotube structure. This approach assumes a chemical strengthening of the
nonbonded interactions with the nanotube
20
and the improvement of the apparent
mechanical connection with the nanotube, for example, by wrapping a large poly-
mer molecule around the nanotube.
21,22
The second approach requires the formation
FIGURE 23.2 Unit vector notational for a graphene sheet.
of a chemical covalent bond between the nanotube and polymer directly, also known
as functionalization. There is reasonable evidence for the presence of such nano-
tube/polymer covalent bonds.
23,24
Some studies have indicated that functionalization
can occur through chemical bonds added to the nanotube sidewalls.
712
Addition of
small organic groups to nanotube sidewalls has been reported via in situ-generated
diazonium compounds
8
and by fluorination of nanotubes
9,10
followed by alkyla-
tion.
11
Functionalization can also occur through oxidizing the nanotube sample to
induce the formation of hydroxyl of carboxylic acid groups at surface defects on
the nanotube.
1318
This type of functionalization can occur with copolymers,
1315
proteins,
16
organosilanes,
17
and metal catalysts.
18
It is known that covalent bonding
affects the elastic mechanical properties of the nanotube itself because the formation
of a chemical bond with carbon atoms in a nanotube interrupts the sp
2
hybridization
of the nanotube, thereby forming a site that is closer to an sp
3
hybridized carbon.
25,26
A recent simulation study by Frankland et al.
27
predicted that for carbon nano-
tube/polyethylene composites there is at least a one-order-of-magnitude increase in
the strength for composites with covalent bonding between the nanotube and adjacent
polymer molecules, relative to systems without the covalent bonds. However, the
hybridization change may weaken the chemical bonds in the vicinity of the function-
alization, and this effect should be manifested as a decrease in the material elastic
constants.
20,21,28
Therefore, a trade-off exists between the effects of increased load
transfer and decreased nanotube properties because of functionalization. The effects
of functionalization are expected to depend on the polymer matrix material.
23.3 MICROMECHANICS
Micromechanical models are used to predict the bulk properties of composite mate-
rials as a function of the properties and geometry of the constituent materials
(Fig. 23.3). In particular, for fiber-reinforced composites, the fiber properties, shape,
FIGURE 23.3 Micromechanical modeling of a composite.
Composite
C
Micromechanics
C
1
Polymer
C
2
Reinforcement
length, volume fraction, orientation, and the matrix properties are used as input into
micromechanical models. Generally in the development and use of micromechanical
models, it is assumed that the constituents are continuous, the properties of which
are mathematical described with continuum mechanics. Micromechanics models for
fiber-reinforced composites can be broken down into two general categories, ana-
lytical and numerical.
23.3.1 ANALYTICAL MICROMECHANICAL MODELS
Many forms of analytical micromechanical models have been developed over the
past 50 years. Important contributions to this effort include the works of Hashin,
29
Hill,
30,31
Hashin and Rosen,
32
Aboudi,
33
Eshelby,
34
Budiansky,
35
Mori and Tanaka,
36
Benveniste,
37
and Halpin and Tsai.
38
Many of these models can be formulated in a
consistent mathematical framework.
Using the direct approach
39
for the estimate of overall properties of heteroge-
neous materials, the volume-averaged stress and strain fields of the composite with
a total of N phases are, respectively,
(23.2)
(23.3)
where V is the stress tensor, H is the strain tensor, c
r
is the volume fraction of
phase r, the overbar denotes a volume-averaged quantity, the subscript r denotes
the phase, and r = 1 is the matrix phase. The constitutive equation for each phase
is given by
(23.4)
where C
r
is the stiffness tensor of phase r. For a composite subjected to homogeneous
elastic strain and electric field boundary conditions, H
0
, it can easily be shown that
.
40
The constitutive equation for the composite can be expressed in terms of
the volume-averaged fields
(23.5)
The volume-average strain field in phase r is
(23.6)
where A
r
is the concentration tensor of phase r, and
(23.7)

c
r r
r 1
N

c
r r
r
N
1

r r r
C

0
C

r r
A
c
r r
r
N
A I
1
where I is the identity tensor. Combining Equation 23.2 and Equation 23.7 yields
the modulus of the composite in terms of the constituent moduli
(23.8)
Various procedures exist for evaluating the concentration tensor. The simplest
approximation is , which is the Voight approximation, also referred to as the
rule-of-mixtures approach. The next simplest approach is the dilute approximation,
for which the interaction between the reinforcement particles is ignored. This results
in the dilute concentration tensor:
(23.9)
where S
r
is the Eshelby tensor for phase r,
34
and is evaluated as a function of the
lengths of the principal axes of the reinforcing phase r, , and the mechanical
properties of the surrounding matrix
(23.10)
Numerous forms of the complete expression for Equation 23.10 are given elsewhere.
34,41
For the MoriTanaka approach, the concentration tensor is
(23.11)
where is the dilute concentration tensor given by Equation 23.9. The
MoriTanaka approach has been shown to be quite accurate, even at larger rein-
forcement volume fractions.
42
In the self-consistent scheme, the concentration tensor
is
(23.12)
where C is the unknown modulus of the composite. Since the moduli of the composite
appear in both Equation 23.8 and 23.12, iterative schemes or numerical techniques
are ultimately required for the prediction of the moduli of composites using the self-
consistent method. This approach results in slow and complicated calculations.
The use of analytical micromechanical models for predicting the properties of
carbon nanotube/polymer composites has been reported in the literature. Pipes and
Hubert
43
developed a model for predicting the properties of arrays of SWNTs that
are assembled in helical arrays of circular cross-section. Anisotropic elasticity was
used to determine the behavior of a layered cylinder with layers of discontinuous
SWNT embedded in a polymeric matrix and with the collimation direction in each
C C C C A +
1 1
2
c
r r r
r
N
( )
A I
r
voi
A I S C C C
r
dil
r r
+
[ ]

1
1
1
1
( )
a
i
r
S C
r
r r r
f a a a
( )
1 1 2 3
, , ,
A A I A
s
MT
s
dil
r r
dil
r
N
c c +
,
,
,
]
]
]
]
1
2
1
A
r
dil
A I S C C C
r
SC
r r
+
[ ]

1
1
( )
layer following a helical path. This model was extended to predict the thermal
expansion properties of such a material
44
and to predict the properties of larger scale
structures composed of multiple helically wound arrays.
45
Odegard et al.
46
used the
MoriTanaka approach to predict the mechanical properties of a (6,6) SWNT/poly-
imide composite with a PmPV nanotube/polyimide interface for various SWNT
volume fractions, lengths, and orientations. A similar approach was used by Odegard
et al.
28
to predict the properties of a (10,10) SWNT/polyethylene composite with
and without chemical functionalization. It was shown that chemical functionaliza-
tion degraded most of the elastic properties of the composites. The models of
Pipes and Hubert
43,45
and Odegard et al.
46
were compared for the same SWNT/
polymer composite material
47
. The two approaches showed agreement in predicted
Youngs moduli for a wide range of SWNT volume fractions. Thostenson and
Chou
48
used the HalpinTsai approach to predict elastic moduli of MWNT/poly-
styrene composites as a function of constituent properties, reinforcement geometry,
and nanotube structure. They demonstrated that the composite elastic properties were
particularly sensitive to the nanotube diameter. Yoon et al.
49
predicted the resonant
frequencies and vibration modes of MWNTs embedded in an elastic medium using
a multiple-elastic beam model. Lagoudas and Seidel
50
used a variety of microme-
chanical techniques, including a two-step MoriTanaka approach and the composite
cylinders approach to examine the properties of carbon nanotube-reinforced epoxy
composites. While the clustering of nanotubes in the composite contributed to some
reduction in composite properties, nanotube cluster misalignment and poor nano-
tubematrix bonding played a much larger role.
23.3.2 NUMERICAL MICROMECHANICAL MODELS
Numerical predictions of bulk properties of carbon nanotube/polymer composites
based on the constituent properties and nanotube geometry and volume fraction can
be achieved with the finite element method (FEM). If performed correctly, the FEM-
based micromechanical approach has the potential to predict bulk properties more
accurately than those predicted with analytical micromechanical techniques. How-
ever, the gain in accuracy when using FEM micromechanical models comes at a
price. Due to the complexity of FEM models and the high costs of FEM software,
implementation of FEM-based micromechanical models can be costly and time
consuming.
The principal advantage of FEM micromechanical models is the direct determi-
nation of the stress and strain distribution in the reinforcement and surrounding
matrix. The careful prediction of these fields results in the accurate estimation of the
bulk properties. The stress and strain fields are determined by discretizing the het-
erogeneous continuum into elements (Fig. 23.4), of which elastic solutions exist for
applied loads and displacements on nodes. The nodes are placed on the corners and
sometimes on the midsides of the element boundaries. As long as the geometry of
the elements (mesh) is not too coarse, the overall predicted properties of FEM models
can be accurately estimated by solving for the simultaneous stress and strain fields
of all of the elements in the model. The pioneering works in the use of FEM-based
micromechanical models include those of Adams,
51
Lin et al.,
52
and Dvorak et al.
53
Several studies have used FEM-based micromechanical models to predict the
properties of carbon nanotube/polymer composites. Liu and Chen
54
and Chen and
Liu
55
used various representative volume element (RVE) geometries to evaluate
the properties of SWNT/polymer composites. These studies demonstrated the
reinforcement potential of carbon nanotubes in nanocomposite materials. Li and
Chou
56
used finite elements to study the interfacial shear stress distribution, stress
concentration in the matrix in the vicinity of carbon nanotube ends, axial stress
profile in the nanotube, and the effect of nanotube aspect ratio on load transfer in
a carbon nanotube/polymer composite. Fisher et al.
57
developed a FEM model to
investigate the effect of fiber waviness on overall mechanical properties of an
effective reinforcing modulus for carbon nanotube/polymer composites. Bradshaw
et al.
58
used FEM models to numerically determine the strain concentration tensor
of a carbon nanotube/polymer composite with wavy carbon nanotubes. These two
studies demonstrated the reduction in mechanical properties that accompanies the
increase in nanotube waviness.
23.4 MOLECULAR MODELS
In the past two decades, computational molecular modeling approaches have
emerged as important tools that can be used to predict atomic structure, vibrational
frequencies, binding energies, heats of reaction, electrical properties, and mechanical
properties of organic and inorganic materials. Unlike the continuum-based micro-
mechanical approaches, no assumptions of a material continuum are made. There-
fore, these methods are ideal in studying materials in which the RVE is on the scale
of nanometers. However, because of the discrete nature of molecular models, only
FIGURE 23.4 Finite element model of a carbon nanotube/polymer composite.
Polymer
Nanotube
a small volume of atoms may be considered in an analysis of a material, which is
a concern in the development of nanostructured materials for structural applications,
such as in the aerospace industry. Furthermore, many molecular modeling
approaches require long simulation times on relatively powerful computers. For the
simulation of carbon nanotube/polymer composites, two main computational molec-
ular modeling techniques are employed: MD and ab initio simulations.
23.4.1 AB INITIO SIMULATIONS
In ab initio simulations of atomic systems, an approximation to the solution of the
Schrdinger equation is performed. The method can be used for systems containing
1 to 40 atoms, depending on the power of the computational equipment available.
Although a wide range of ab initio methods have been developed, the most com-
monly used approach is the HartreeFock method. More detailed discussions of ab
initio simulations can be found elsewhere.
59
There are many procedures that can be followed in ab initio simulations. Three
commonly used procedures are single point calculations, geometry optimization
calculations, and frequency calculations. In single point calculations, the energy and
wave function are calculated for a fixed geometry. This approach is sometimes
employed as a preliminary step in a more detailed study on a molecular structure
or after a geometry optimization. In geometry optimization calculations, the wave
function and energy of a molecule are calculated at a starting geometry and subse-
quent geometries with lower energy levels. The process is repeated until a minimized
energy geometry is achieved. Various procedures exist for establishing the geome-
tries for every calculation step. Frequency calculations are used to predict the
frequencies and the infrared and Raman intensities of a molecular system.
Ab initio studies of carbon nanotube/polymer composites in the literature are
limited. Mylvaganam and Zhang
60
studied the bonding between polyethylene and
carbon nanotubes. The study predicted that the bonding is energetically favorable
and that the tubes of smaller diameters have higher binding energies. Jaffe
61
exam-
ined the bonding of fluorine atoms and nonplanar carbon geometries corresponding
to (10,10), (5,5), and (16,0) carbon nanotubes. It was shown that zigzag nanotubes
form more stable fluorination products than armchair tubes of comparable diameter.
Bauschlicher
62,63
studied the binding energies of hydrogen to (10,0), (9,0), and (5,5)
carbon nanotubes. Hydrogen coverages of 50% and 100% were applied onto the
carbon nanotubes. Bauschlicher
64
studied the bonding of hydrogen and fluorine
atoms to the sides of (10,0) carbon nanotubes. The fluorine atoms appeared to favor
bonding next to existing fluorine atoms, while the addition of two or four hydrogen
atoms to the nanotubes was shown to be endothermic.
23.4.2 MOLECULAR DYNAMICS
Molecular dynamics is a computational technique in which a time evolution of a set
of interacting atoms is followed by integrating their equations of motion. The motion
of the atoms is described by Newtons law:
(23.13) F a
i i i
m
where F
i
is the force vector acting on atom i, and m
i
and a
i
are the mass and acceleration
vectors, respectively, of atom i in a system with N atoms. The forces acting on each
atom are due to the interactions with the other atoms. A trajectory is calculated in a
6N-dimensional phase space (three positions and three momenta components for each
atom). Since MD is a statistical mechanics method, a set of configurations distributed
according to a statistical distribution function (statistical ensemble) is obtained. Typical
MD simulations are performed on systems containing thousands of atoms and for
simulation times ranging from picoseconds to nanoseconds.
The physical quantities of the system are represented by averages over config-
urations distributed according to the chosen statistical ensemble. A trajectory
obtained with MD provides such a set of configurations. Therefore the computation
of a physical quantity is obtained as an arithmetic average of the instantaneous
values. Statistical mechanics is the link between the microscopic behavior and
thermodynamics. Thus the atomic system is expected to behave differently for
different pressures and temperatures.
The interactions of the particular atom types are described by the total potential
energy of the system as a function of the positions of the individual atoms at a
particular instant in time:
(23.14)
where r
i
represents the coordinates of atom i in a system of N atoms. The potential
equation is invariant to coordinate transformations, and is expressed in terms of the
relative positions of the atoms with respect to each other, rather than from absolute
coordinates. The forces on atom i are determined from the gradient of the potential
with respect to the atomic displacements
(23.15)
where is the gradient operator. The total energy of the system is
(23.16)
where K is the instantaneous kinetic energy, which is dependent on the thermody-
namic condition of the system.
Numerous options exist for the selection of the potential, V. For systems that
consist of only carbon or hydrogen atoms, such as carbon nanotube/polyethylene
composites, the Brenner potential
65
has been extensively used.
27,66,67
The basic con-
cept of the Brenner potential is that the strength of a bond between two atoms is
not constant, but depends on local conditions. The potential is expressed as
(23.17)
where the summations are performed over the bonds in the system, r
ij
is the distance
between atoms i and j, V
R
(r
ij
) and V
A
(r
ij
) are repulsive and attractive interaction terms,
V V
i N
( , , ) r r
F
i N
V ( ) r r
1
,...,
E K V +
V V r B V r
R ij ij A ij
j i i

>

[ ( ) ( )]
( )
respectively, and B
ij
represents a many-body coupling between the bond from atom
i to atom j and the local environment of atom i. For organic material systems that
consist of more than just carbon and hydrogen atoms, a simple yet effective approach
is to describe the potential with molecular mechanics.
6872
With the molecular
mechanics approach, the potential energy of the system is described by the sum of
the individual energy contributions of each degree of freedom for each of the N
atoms (Fig. 23.5). The form of the individual energy contributions is often referred
to as a force field. For example, for the force field of Cornell et al.,
68
the potential
energy is
(23.18)
where K
r
is the bond stretching force constant, r is the distance between atoms, r
eq
is the equilibrium spacing between atoms, K
T
is the bond-angle bending force
constant, T is the bond angle, T
eq
is the equilibrium bond angle, V
n
is the torsion
force constant, J is the phase offset, n is the periodicity of the torsion, A
IJ
and B
IJ
are van der Waals force constants between nonbonded atoms I and J, and r
IJ
is the
distance between atoms I and J. In Equation 23.18, the summations occur over all
FIGURE 23.5 Molecular mechanics degrees of freedom.
Stretching
Angle bending
Torsion
Vander Waals
Improper torsion
r
s
h
h
r
V K r r K
V
r eq
bonds
eq
angles
n
+
+ +

( ) ( )
[
2 2
2
1

ccos( )] n
A
r
B
r
dihedrals
IJ
IJ
IJ
IJ
+
,
,
]
]

12 6 ]]
<
I J
of the corresponding interactions in the molecular system. The van der Waals term
in Equation 23.18 is also known as the LennardJones potential. Examples of force
constants are shown in Table 23.1 for aliphatic carbon atoms. More details on MD
simulation technique can be found elsewhere.
73
Numerous studies have addressed the simulation of carbon nanotube/polymer
composites using molecular mechanics and MD techniques. Lordi and Yao
22
used
molecular mechanics simulations to calculate sliding frictional stresses between
carbon nanotubes and a range of polymer substrates. It was determined that fric-
tional forces and helical polymer conformations played minor and major roles
respectively in determining the interfacial strength. Frankland et al.
27
used MD
simulations to examine the influence of functionalization between a SWNT and
polyethylene matrix on shear strengths and critical lengths required for load
transfer. The simulations predicted that shear strengths and critical lengths required
for load transfer increase and decrease, respectively, when the nanotubes are
functionalized to the matrix. Frankland and Harik
74
studied the pull-out of SWNT
from a polyethylene matrix to characterize an interfacial friction model. A periodic
variation in the nanotube displacements and velocities was observed, which cor-
responded to the atomic structure of the carbon nanotube. Frankland et al.
66
predicted stressstrain curves of SWNT/polyethylene composites with infinite and
finite-length nanotubes. The curves were subsequently compared to micromechan-
ical predictions. Hu and Sinnott
75
and Hu et al.
76
investigated the effects of
chemical functionalization of carbon nanotube/polystyrene composite on the com-
posite toughness. The chemical modification of the carbon nanotubes during
polyatomic ion beam deposition was detailed. Jaffe and coworkers
7779
examined
the interaction of carbon nanotubes, water droplets, and aqueous solutions. Radial
density profiles and radial hydrogen bond distributions were predicted. Liang et al.
80
demonstrated the attractive interaction between (10,10) SWNTs and an epoxy poly-
mer matrix. The results indicated that the aromatic ring structures of the polymer
molecules have a tendency to align with the aromatic structures in the SWNTs. Wei
et al.
81
showed that the addition of nanotubes to a polyethylene matrix increased
the glass transition temperature, thermal expansion, and diffusion coefficients of
TABLE 23.1
Force Constants for Aliphatic Carbon Atoms
Bond Parameters
CC bond stretch K
r
310.0 kcal/(mole
2
)
r
eq
1.526
CCC bond angle K
T
40.0 kcal/(mole rad
2
)
T
eq
109.5q
CCCC torsion V
n
/2 1.4 (kcal/mole)
J 0q
n 3
CC van der Waals interaction R
*
1.908
H 0.1094 (kcal/mole)
the polymer. Griebel and Hamaekers
67
determined stressstrain curves of nano-
tube/polyethylene composites. Good agreement with rule-of-mixture microme-
chanical models was found. Odegard et al.
82
predicted the mechanical properties
of cross-linked nanotubes for various SWNT cross-link densities. While the cross-
linking significantly increased the transverse shear moduli of the nanotube arrays
by a factor of up to 30, the corresponding degradation of the nanotube caused an
overall decrease in the axial properties of 98%.
23.5 EXAMPLE: SWNT/POLYIMIDE COMPOSITE
For this example, the constitutive properties of a nanotube/polyimide composite with
a poly(m-phenylenevinylene) (PmPV) interface were predicted based on the method
developed by Odegard et al.
46
This method relies on an equivalent continuum mod-
eling technique
83
used to predict the bulk mechanical behaviors of nanostructured
materials. In summary, the method consists of four major steps: establishing repre-
sentative volume elements (RVEs) for the molecular and equivalent continuum mod-
els, establishing a constitutive relationship for the equivalent continuum model, deriv-
ing and equating potential energies of deformation for both models subjected to
identical boundary conditions, and using traditional micromechanics techniques to
determine larger scale properties of the composite. Each step of the modeling is
described in the next section. Further details of the modeling are discussed else-
where.
46
23.5.1 MODELING PROCEDURE
In order to establish the RVE of the molecular model, and thus the equivalent
continuum model, the molecular structure was first determined. An MD simulation
was used to generate the equilibrium structure of a nanotube/polyimide composite,
which consisted of a (6,6) single-wall nanotube and five PmPV oligomers, each ten
repeating units in length. The initial structure was constructed by placing the nano-
tube at the center of the MD cell, and by inserting the PmPV molecules at random,
nonoverlapping positions within the MD cell. The molecular structure for a single
time increment for an equilibrated system was used as the RVE of the molecular
model (Fig. 23.6). The RVE of the equivalent continuum model was chosen to be
a solid cylinder of the same dimensions as the molecular model RVE. The equivalent
continuum model will be referred to as the effective fiber for the remainder of this
example discussion.
The second step in the modeling technique involved establishing a constitutive
equation for the effective fiber. It was assumed that the effective fiber had a linear
elastic constitutive behavior. Examination of the molecular model (Fig. 23.6)
revealed that it was accurately described as having transversely isotropic symmetry,
with the plane of isotropy perpendicular to the long axis of the nanotube. There are
five independent elastic constants required to determine the entire set of elastic
constants for a transversely isotropic material.
In the third step of the modeling approach, the potential energies of deformation
for the molecular model and effective fiber were derived and equated for identical
loading conditions. Five independent elastic constants were determined by employ-
ing a technique adapted from the approach of Hashin and Rosen.
32
These elastic
constants were the transverse shear modulus, , transverse bulk modulus, , lon-
gitudinal shear modulus, , longitudinal Youngs modulus, , and the longitudinal
elastic stiffness component, , where the superscript f denotes effective fiber. The
longitudinal properties were associated with the direction parallel to the longitudinal
axis of the nanotube, and the transverse properties were associated with the trans-
verse plane of isotropy (Fig. 23.6).
The bonded and nonbonded interactions of the atoms in a polymeric molecular
structure were described by using a force field similar to that shown in Equation 23.18.
Each of the five elastic constants of the effective fiber was determined from a
single boundary condition applied to both molecular and effective fiber models.
Therefore, for each applied boundary condition, one elastic constant of the effec-
tive fiber was uniquely determined. For this particular study, an intermediate step
was used to equate the molecular and effective fiber models. An equivalent truss
model was used to approximate the energy of the molecular model. The calculated
values of the five independent parameters for the effective fiber are listed in
Table 23.2.
Overall constitutive properties of the nanocomposite with randomly oriented and
unidirectional nanotubes were determined with the micromechanical-based
MoriTanaka method
36,37
by using the mechanical properties of the effective fiber
and the bulk polymer matrix material (Fig. 23.3). The layer of polymer molecules
near the polymer/nanotube interface was included in the effective fiber, and it was
assumed that the matrix polymer surrounding the effective fiber had mechanical
FIGURE 23.6 Equivalent continuum model of the nanotube/PmPV composite.
x
3
x
2
Molecular
model
Equivalent-
continuum model
Polymer atoms Nanotube atoms
x
2
x
1
G
T
f
K
T
f
G
L
f
Y
L
f
C
L
f
properties equal to those of the bulk polymer. Because the bulk polymer molecules
and the polymer molecules included in the effective fiber are physically entangled,
perfect bonding between the effective fiber and the surrounding polymer matrix was
assumed. The bulk polymer matrix material had a Youngs modulus and Poissons
ratio of 3.8 GPa and 0.4, respectively.
The maximum nanotube volume fraction that can be obtained with this approach
is limited by the maximum effective fiber volume fraction that can be used in the
micromechanics analysis. For a hexagonal packing arrangement, the maximum
effective fiber volume fraction is 90.7%. While the nanotube and effective fiber
lengths are equal, the nanotube volume fraction is 34% of the effective fiber volume
fraction if it is assumed that the nanotube volume is a hollow cylinder with a wall
thickness equal to the interatomic spacing of graphene sheets (0.34 nm).
23.5.2 RESULTS AND DISCUSSION
In this section, the moduli of the effective-fiber/polyimide composite are presented
in terms of nanotube length, volume fraction, and orientation. Figure 23.7 is a plot
of the calculated longitudinal Youngs modulus, E
L
, and longitudinal shear modulus,
G
L
, for the random and aligned composites as a function of nanotube length, for a
1% nanotube volume fraction. These quantities were calculated from the elastic
stiffness tensor of the composite, C, by using Equation 23.8 for the composite. The
results indicate an approximately 55% increase in the shear modulus of the randomly
oriented nanotube composite in the range of nanotube lengths between 0 to 200 nm,
with a significant change in the slope between nanotube lengths of 50 and 100 nm.
Conversely the calculated longitudinal shear modulus for the aligned nanotube
composite was constant for the given range of nanotube length. Therefore, under
these conditions, increasing the degree of alignment resulted in a decrease in shear
modulus.
From Fig. 23.7 it can be seen that alignment of the nanotubes results in nearly
a 300% increase in the longitudinal Youngs modulus for the composite with nano-
tubes that are 200 nm long with respect to the composite with short nanotubes. Unlike
the case for shear modulus, an increase in the degree of alignment resulted in an
increase in the longitudinal Youngs modulus. A significant decrease in the slope of
the Youngs modulus curves occurs between nanotube lengths of 60 to 80 nm.
TABLE 23.2
Elastic Properties of Effective Fiber
G
T
f
4 4 . GPa
K
T
f
9 9 . GPa
G
L
f
27 0 . GPa
Y
L
f
450 4 . GPa
C
L
f
457 6 . GPa
The longitudinal Youngs modulus of the aligned composite is plotted in
Fig. 23.8 as a function of nanotube volume fraction for nanotubes that are 10, 50,
and 500 nm long. The Youngs modulus increased with an increase in volume
fraction, with the most pronounced rate of increase associated with nanotubes of
length 50 nm or greater. The dependence of the longitudinal Youngs modulus on
the nanotube volume fraction became more linear as the nanotube length increased.
This dependence is expected because of the well known effect of the increase in
load transfer with subsequent increases in reinforcement length and volume fraction.
84
FIGURE 23.7 Modulus of SWNT/polyimide composite material versus nanotube length for
a 1% nanotube volume fraction.
FIGURE 23.8 Longitudinal Youngs modulus of aligned SWNT/polyimide composite versus
nanotube volume fraction.
0
2
4
6
8
10
12
14
16
0 50 100 150 200
Nanotube length (nm)
M
o
d
u
l
u
s

(
G
P
a
)
Longitudinal Youngs modulus
Longitudinal shear modulus
Aligned
Random
Aligned
Random
0
50
100
150
200
250
300
350
0 10 15 20 25
Nanotube volume fraction (%)
L
o
n
g
i
t
u
d
i
n
a
l

Y
o
u
n
g
s

m
o
d
u
l
u
s

(
G
P
a
)
L = 10 nm
L = 50 nm
L = 500 nm
5
For 500-nm-long aligned nanotubes, the longitudinal Youngs modulus of a com-
posite with a 25% nanotube volume fraction was about 85 times larger than the
Youngs modulus of the unreinforced resin.
Figure 23.9 is a plot of the Youngs modulus and the shear modulus for the
random composite as a function of nanotube volume fraction, for three nanotube
lengths. In general, an increase in nanotube volume fraction resulted in increased
moduli values. For both the Youngs and the shear moduli, increasing the volume
fraction for the nanotubes of length near 10 nm provided little to no improvement
in stiffness. However, for nanotubes between 50 nm to 200 nm, equivalent stiffness
can easily be obtained by trading off a decrease in nanotube length for a small (2u or
less) change in volume fraction.
23.6 EXAMPLE: SWNT/POLYETHYLENE COMPOSITE
For this example, stressstrain curves of two unidirectional polyethylene/NT compos-
ites were generated from MD simulations.
66
The first composite contained an infinitely
long SWNT, and the second composite contained a short SWNT. In the following
discussion, details of the composite structure and MD simulations are presented. A
detailed description of the computation of stress from molecular force fields follows.
Finally the stressstrain curves generated from MD simulation are presented for
longitudinal and transverse loading conditions for each of the two composites
considered.
23.6.1 MD SIMULATIONS
The unidirectional SWNT/polymer composites considered in this example contained
long and short SWNTs, as shown in Fig. 23.10. The dashed boxes in Fig. 23.10
enclose a RVE that was simulated by MD. The SWNT composite in Fig. 23.10(a)
FIGURE 23.9 Modulus of random SWNT/polyimide composite material versus nanotube
volume fraction.
0
5
10
15
20
0 2 6
Nanotube volume fraction (%)
M
o
d
u
l
u
s

(
G
P
a
)
E
E
E
G
G
G
E - Youngs modulus
G - Shear modulus
L = 10 nm
L = 50 nm
L = 200 nm
1 5 4 3 7
contained a periodically replicated (10,10) SWNT that spanned the length of the
simulation cell. In this composite, the SWNT was embedded in an amorphous
polyethylene matrix, which was represented by beads of united atom CH
2
units.
Specifically the polyethylene matrix had eight chains of 1095 CH
2
units. The short-

SWNT composite, shown in Fig. 23.10(b), contained a 6-nm capped (10,10) SWNT
that was approximately half the length of the simulation cell. The SWNT caps each
consisted of one-half of a C
240
molecule. In this composite, the amorphous polyeth-
ylene matrix contained eight chains of 1420 CH
2
units. The overall dimensions

for the unit cells of each composite in the MD simulation were approximately
5 u 5 u 10 nm. Periodic boundary conditions were used to replicate the cell in all
three dimensions. For comparison to the SWNT/polyethylene composites, an equiv-
alent-sized block of amorphous polyethylene without a SWNT was also simulated.
For each structure, a polyethylene density of 0.71 g/cm
3
was used.
In the MD simulation, the van der Waals interfacial interaction between the poly-
mer and the SWNT was modeled with the LennardJones potential Equation 23.18).
The polyethylene chains were simulated with a molecular mechanics force field
adapted from the literature.
66
Specifically the CH
2
units of the polyethylene chains

were separated by bond lengths of 0.153 nm by using the SHAKE algorithm, a
constraint dynamics method that constrains the bond length within a user-defined
tolerance.
73
Angle-bending forces were modeled with a harmonic valence angle poten-
tial having an equilibrium angle of 112.813
o
and a barrier of 520 kJ/mol. A torsional
potential was used for the torsion angle around the CH
2
-CH
2
bond.
66
The Len-
nardJones potential was also used to describe nonbonding interactions between
CH
2
units in either the same chain or between different chains. The SWNT was
modeled with a many-body bond-order potential developed for carbon.
85
This carbon
potential was parameterized for CC bonds of lengths up to 0.17 nm, which was
within the magnitude of the strain applied to the composites in the present work. All
simulations were carried out at 300 K, with 2 femtosecond time steps.
FIGURE 23.10 Schematic of polymer nanocomposites filled with long and short carbon
nanotubes.
Nanotubes
Periodic cells
6 nm
x
1
x
2
(a) (b)
5 nm
10 nm
Polymer
23.6.2 STRESSSTRAIN CURVES FROM SIMULATION
Stressstrain curves were generated for the long and short SWNT composites
and for the pure polymer via MD simulation. For both composite configurations,
the longitudinal (parallel to the SWNT axis) and transverse responses were
simulated. The prescription of strain and calculation of stresses is explained in
this section.
For each increment of applied deformation, a uniform strain was prescribed on
the entire MD model. For the longitudinal and transverse deformations, pure states
of strain, H
11
and H
22
, respectively, were initially applied (Fig. 23.11). The application
of strain was accomplished by uniformly expanding the dimensions of the MD
cell in the direction of the deformation and rescaling the new coordinates of the
atoms to fit within the new dimensions. After this initial deformation, the MD
simulation was continued and the atoms were allowed to equilibrate within the
new MD cell dimensions. This process was carried out for the subsequent incre-
ments of deformation. The applied strain increment in both the longitudinal and
transverse directions was 2%, and was applied in two equal increments of 1%.
After each 2% increment of strain, the system was relaxed for 2 ps, and then the
stress on the system was averaged over an interval of 10 ps. For each composite
configuration, six increments of 2% strain were applied up to a total of approx-
imately 12% over a period of 72 ps. The corresponding strain rate was 1.0 u 10
10
s
1
. This high strain rate is inherent to MD simulation, which includes dynamic
information usually on picosecond-to-nanosecond timescales.
FIGURE 23.11 Definitions of the strains applied to the composites.
x
2
x
1
x
3
11
22
In general, the stress in a solid (or a group of interacting particles in the form of
a solid) is defined as the change in the internal energy (in the thermodynamic sense)
with respect to the strain per unit volume. For example, at the continuum level, the
stress tensor, V
ij
, for a linear-elastic material is
86
(23.19)
where V is the volume of the solid, E is the total internal energy, H
ij
is the strain
tensor, and the subscript S denotes constant entropy. When the internal energy is
equal to the strain energy of the solid, Hookes law may be derived from
Equation 23.19. Furthermore, if the strain energy is expressed in terms of an applied
force acting over the surface area of a solid, then a more familiar form of stress as
force per unit area is derived.
At the atomic level, the total internal energy given in Equation 23.19 can be
expressed as the summation of the energies of the individual atoms, E
D
, that compose
the solid:
(23.20)
where for each atom D, T
D
is the kinetic energy, U
D
is the potential energy, M
D
is
the mass, v
D
is the magnitude of its velocity, and )
D
(r) is the potential energy at
the atom location r. Using a Hamiltonian that is based on these individual energy
contributions, E
D
, it has been shown that the stress contribution, , for a given
atom is
(23.21)
where V
D
is the atomic volume of atom D, is the i-component of the velocity of
atom D, Q
j
D
is the j-component of the velocity of atom D, is the i-component of
the force between atoms D and E obtainable from the derivative of the potential
)(r), and is the j-component of the separation of atoms D and E.
87,88
These
parameters are shown in Fig. 23.12 as well.
The stresses that were used to generate the stress-strain curves for the SWNT
composites were average atomic stresses for the volume of the model. Therefore, the
stress components of each model were calculated for each strain increment by using
(23.22)
ij
ij
S
V
E
j
(
,
\
,
(
1
E T U M v r

+ +
1
2
2
( ) ( )
ij
ij i j i j
V
M v v F r +
j
(
,
,
\
,
(
(
1
v
i
F
i
r
j

ij i j i j
V
M v v F r +
j
(
,
,
\
,
(
(

1
where V is the volume of the MD model and The stress calculated with
Equation 23.22 was then averaged over time via the MD simulation.
23.6.3 RESULTS
The stressstrain curve of the long SWNT composite under longitudinal loading is
shown in Fig. 23.13. The stressstrain curve of the polymer is also shown in
Fig. 23.13. A large increase in the slope of the stressstrain curve is observed upon
reinforcement of the polymer material. The stressstrain curve of the short SWNT
FIGURE 23.12 Parameters used to compute stresses in the simulation.
FIGURE 23.13 Longitudinal stressstrain relation of long NT composite and polymer matrix
without nanotubes.
n
= 1
= 2
= 3
F
1
F
2
F
3
r
1
r
3
r
2
x
1
x
3
x
2
V V

.
0
1
2
3
4
0 6 10
Strain (%)
S
t
r
e
s
s

(
G
P
a
)
Long NTs
No NTs
4 2 8
subjected to longitudinal loading conditions is plotted with the simulated polymer
stressstrain curve in Fig. 23.14. Almost no enhancement relative to the polymer is
observed in the stressstrain curve of the composite with the short SWNT. The
stressstrain curves of both composites subjected to transverse loading conditions are
compared with each other and the polymer in Fig. 23.15. All three MD-generated
curves are similar. For the composite with the long SWNT, the stress after loading
FIGURE 23.14 Longitudinal stressstrain relation for short NT-filled composite compared
with stressstrain behavior of the polymer matrix without nanotubes.
FIGURE 23.15 Transverse stressstrain curves of both long and short NT-filled composites
compared with polymer matrices without nanotubes.
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0 8 10
Strain (%)
S
t
r
e
s
s

(
G
P
a
)
12
No NTs
Short NTs
6 4 2
0.00
0.05
0.10
0.15
0.20
0.25
0 10
Strain (%)
S
t
r
e
s
s

(
G
P
a
)
12
Long NTs
Short NTs
No NTs
8 6 4 2
in the transverse direction is approximately 30 times lower than the stress levels after
loading to the same strain level in the longitudinal direction (Fig. 23.13). This
difference in longitudinal and transverse behavior illustrates the anisotropy of the
composite.
This chapter has discussed some of the fundamental concepts of the modeling of
carbon nanotube/polymer composites. Four principal modeling tools have been
described: analytical micromechanics, numerical micromechanics, ab initio simu-
lations, and MD. A comprehensive summary of recent work performed using these
four modeling approaches has been presented, along with two in-depth examples
of the modeling of carbon nanotube/polyimide and carbon nanotube/polyethylene
composites.
The principal difficulty in modeling nanostructured materials is the selection of
the correct (or combination) of modeling tools. Micromechanical models are conve-
nient; however, their fundamental assumption is that a continuum exists in the RVE.
For a nanostructured material, the RVE is best described as a lattice structure of atoms
and molecules. Therefore micromechanical models alone are not expected to accu-
rately predict the behavior of nanotube/polymer composites. Finite element models
are more difficult to develop and run than the analytical micromechanical models,
but can be more accurate. However, finite element models also assume the presence
of a continuum on the nanometer-length scale. Ab initio models operate at a first-
principles level, and thus predict atomic structures in a very accurate manner. How-
ever, the computational difficulty associated with them limits the model size to a very
small set of atoms. MD models are capable of modeling larger molecular systems
than ab initio simulations; however, information about atomic properties is lost in
the development of the atomic potentials. A combination of modeling tools can be
successfully used to predict the properties of carbon nanotube/polymer composites.
The first example (Section 23.5) used MD to establish the atomic structure of the
nanotube and polymer molecules in the RVE, and subsequently used micromechanics
to predict mechanical properties of the composite at a larger length scale. Therefore,
the choice of a modeling tool (or combination of tools) for carbon nanotube/polymer
composites should be based on size of the RVE, the computational resources available,
and the structure of the material at length scales larger than the RVE.
PROBLEMS
1. How does the use of carbon nanotubes as reinforcement in composite
materials improve the mechanical and electrical properties over those of
the pure polymer resin?
2. In what ways do nanotubes offer advantages as polymer reinforcements
over graphite and glass fibers?
3. What information is lost when continuum mechanics-based models are
used to describe the behavior of an atomic structure?
4. What limits do molecular modeling techniques (MD and ab initio simu-
lations) have in predicting behavior of heterogeneous materials?
REFERENCES
1. Hull, D. and Clyne, T.W., An Introduction to Composite Materials, Cambridge University
Press, Cambridge, 1996.
2. Kroto, H.W., Heath, J.R., OBrien, S.C., Curl, R.F., and Smalley, R.E., C60: Buck-
minsterfullerene, Nature, 318, 162, 1985.
4. Thostenson, E.T., Ren, Z., and Chou, T.W., Advances in the science and technology
of carbon nanotubes and their composites: a review, Compos. Sci. Technol., 61, 1899,
2001.
5. Lau, K.T. and Hui, D., The revolutionary creation of new advanced materialscarbon
nanotube composites, Compos. Part B, 33, 263, 2002.
6. Sun, Y.P., Fu, K., Lin, Y., and Huang, W., Functionalized carbon nanotubes: properties
and applications, Accounts Chem. Res., 35, 1096, 2002.
7. Chen, J., Hamon, M.A., Hu, H., Chen, Y., Rao, A.M., Eklund, P.C., and Haddon,
R.C., Solutions properties of single-walled carbon nanotubes, Science, 282, 95, 1998.
8. Bahr, J.L. and Tour, J.M., Highly functionalized carbon nanotubes using in situ
generated diazonium compounds, Chem. Mater., 13, 3823, 2001.
9. Michelson, E.T., Huffman, C.B., Rinzler, A.G., Smalley, R.E., Hauge, R.H., and
Margrave, J.L., Fluorination of buckytubes, Chem. Phys. Lett., 296, 188, 1998.
10. Michelson, E.T., Chiang, I.W., Zimmerman, J.L., Boul, P.J., Lozano, J., Liu, J.,
Smalley, R.E., Hauge, R.H., and Margrave, J.L., Solvation of fluorinated single-wall
carbon nanotubes in alcohol solvents, J. Phys. Chem. B, 103, 4318, 1999.
11. Boul, P.J., Liu, J., Michelson, E.T., Huffman, C.B., Ericson, L.M., Chiang, I.W.,
Smith, K.A., Colbert, D.T., Hauge, R.H., Margrave, J.L., and Smalley, R.E., Revers-
ible sidewall functionalization of buckytubes, Chem. Phys. Lett., 310, 367, 1999.
12. Chen, Y., Haddon, R.C., Fang, S., Rao, A.M., Eklund, P.C., Lee, W.H., Dickey, E.C.,
Grulke, E.A., Pendergrass, J.C., Chavan, A., Haley, B.E., and Smalley, R.E., Chemical
attachment of organic functional groups to single-walled carbon nanotube material,
J. Mater. Res., 13, 2423, 1998.
13. Jin, Z., Sun, X., Xu, G., Goh, S.H., and Ji, W., Non-linear optical properties of some
polymer/multi-walled carbon nanotube composites, Chem. Phys. Lett., 318, 505,
2000.
14. Hill, D.E., Lin, Y., Rao, A.M., Allard, L.F., and Sun, Y.P., Functionalization of carbon
nanotubes with polystyrene, Macromolecules, 35, 9466, 2002.
15. Lin, Y., Rao, A.M., Sadanadan, B., Kenik, E.A., and Sun, Y.P., Functionalizing
multiple-walled carbon nanotubes with aminopolymers, J. Phys. Chem. B, 106, 1294,
2002.
16. Huang, W., Taylor, S., Fu, K., Lin, Y., Zhang, D., Hanks, T.W., Rao, A.M., and Sun,
Y.P., Attaching proteins to carbon nanotubes via diimide-activated amidation, Nano
Lett., 2, 311, 2002.
17. Velasco-Santos, C., Martinez-Hernandez, A.L., Lozada-Cassou, M., Alvarex-Castillo,
A., and Castano, V.M., Chemical functionalization of carbon nanotubes through an
organosilane, Nanotechnology, 13, 495, 2002.
18. Banerjee, S. and Wong, S.S., Structural characterization, optical properties, and
improved solubility of carbon nanotubes functionalized with Wilkinsons catalyst,
J. Am. Chem. Soc., 124, 8940, 2002.
19. Sinnott, S.B., Chemical functionalization of carbon nanotubes, J. Nanosci. Nanotechnol.,
2, 113, 2002.
20. Chen, R.J., Zhang, Y., Wang, D., and Dai, H., Noncovalent sidewall functionalization
of single-walled carbon nanotubes for protein immobilization, J. Am. Chem. Soc.,
123, 3838, 2001.
21. Star, A., Stoddart, J.F., Steuerman, D., Diehl, M., Boukai, A., Wong, E.W., Yang, X.,
Chung, S., Choi, H., and Heath, J.R., Preparation and properties of polymer-wrapped
single-walled carbon nanotubes, Angew. Chemie Int. Ed., 40, 1721, 2001.
22. Lordi, V. and Yao, N., Molecular mechanics of binding in carbon-nanotubepolymer
composites, J. Mater. Res., 15, 2770, 2000.
23. Jia, Z., Wang, Z., Xu, C., Liang, J., Wei, B., Wu, D., and Zhu, S., Study on poly(methyl
methacrylate)/carbon nanotube composites, Mater. Sci. Engin. A, 271, 395, 1999.
24. Wagner, H.D., Lourie, O., Feldman, Y., and Tenne, R., Stress-induced fragmentation
of multiwall carbon nanotubes in a polymer matrix, Appl. Phys. Lett., 72, 188, 1998.
25. Georgakilas, V., Kordatos, K., Prato, M., Guldi, D.M., Holzinger, M., and Hirsch,
A., Organic functionalization of carbon nanotubes, J. Am. Chem. Soc., 124, 760, 2002.
26. Georgakilas, V., Voulgaris, D., Vazquez, E., Prato, M., Guldi, D.M., Kukovecz, A.,
and Kuzmany, H., Purification of HiPCO carbon nanotubes via organic functional-
ization, J. Am. Chem. Soc., 124, 14318, 2002.
27. Frankland, S.J.V., Caglar, A., Brenner, D.W., and Griebel, M., Molecular simulation
of the influence of chemical cross-links on the shear strength of carbon nano-
tubepolymer interfaces, J. Phys. Chem. B, 106, 3046, 2002.
28. Odegard, G.M., Frankland, S.J.V., and Gates, T.S., The effect of chemical function-
alization on mechanical properties of nanotube/polymer composites, in 44th
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Con-
ference, Norfolk, VA, 2003, p. 1701.
29. Hashin, Z., The elastic moduli of heterogeneous materials, J. Appl. Mech., 29, 143,
1962.
30. Hill, R., Theory of mechanical properties of fibre-strengthened materials I. Elastic
behavior, J. Mech. Phys. Solids, 12, 199, 1964.
31. Hill, R., A self-consistent mechanics of composite materials, J. Mech. Phys. Solids,
13, 213, 1965.
32. Hashin, Z. and Rosen, B.W., The elastic moduli of fiber-reinforced materials, J. Appl.
Mech., 31, 223, 1964.
33. Aboudi, J., Mechanics of Composite Materials: A Unified Micromechanical
Approach, Elsevier, Amsterdam, 1991.
34. Eshelby, J. D., The determination of the elastic field of an ellipsoidal inclusion, and
related problems, Proc. Roy. Soc. London, Series A, 241, 376, 1957.
35. Budiansky, B., On the elastic moduli of some heterogeneous materials, J. Mech. Phys.
Solids, 13, 223, 1965.
36. Mori, T. and Tanaka, K., Average stress in matrix and average elastic energy of
materials with misfitting inclusions, Acta Metal., 21, 571, 1973.
37. Benveniste, Y., A new approach to the application of MoriTanaka's theory in com-
posite materials, Mech. Mater., 6, 147, 1987.
38. Halpin, J.C. and Tsai, S.W., Environmental Factors in Composites Design, AFML-
TR-67-423.
39. Hill, R., Elastic properties of reinforced solids: Some theoretical principles, J. Mech.
Phys. Solids, 11, 357, 1963.
40. Hashin, Z., Theory of Fiber Reinforced Materials, NASA CR-1974.
41. Mura, T., Micromechanics of Defects in Solids, Martinus Nijhoff, The Hague, 1982.
42. Benedikt, B., Rupnowski, P., and Kumosa, M., Visco-elastic stress distributions and
elastic properties in unidirectional composites with large volume fractions of fibers,
Acta Mater., 51, 3483, 2003.
43. Pipes, R.B. and Hubert, P., Helical carbon nanotube arrays: mechanical properties,
Compos. Sci. Technol., 62, 419, 2002.
44. Pipes, R.B. and Hubert, P., Helical carbon nanotube arrays: thermal expansion, Com-
pos. Sci. Technol., 63, 1571, 2003.
45. Pipes, R.B. and Hubert, P., Scale effects in carbon nanostructures: self-similar anal-
ysis, Nano Lett., 3, 239, 2003.
46. Odegard, G.M., Gates, T.S., Wise, K.E., Park, C., and Siochi, E.J., Constitutive modeling
of nanotube-reinforced polymer composites, Compos. Sci. Technol., 63, 1671, 2003.
47. Odegard, G.M., Pipes, R.B., and Hubert, P., Comparison of two models of SWCN
polymer composites, Compos. Sci. Technol., 64, 1011, 2004.
48. Thostenson, E.T. and Chou, T.W., On the elastic properties of carbon nanotube-based
composites: modeling and characterization, J. Phys. D: Appl. Phys., 36, 573, 2003.
49. Yoon, J., Ru, C.Q., and Mioduchowski, A., Vibration of an embedded multiwall
carbon nanotube, Compos. Sci. Technol., 63, 1533, 2003.
50. Lagoudas, D.C. and Seidel, G.D., Effective Elastic Properties of Carbon Nanotube
Reinforced Composites, in 45th AIAA/ASME/ASCE/AHS/ASC Structures, Structural
Dynamics, and Materials Conference, Palm Springs, CA, 2004.
51. Adams, D.F., Inelastic analysis of a unidirectional composite subjected to transverse
normal loading, J. Compos. Mater., 4, 310, 1970.
52. Lin, T.H., Salinas, D., and Ito, Y.M., Elastic-plastic analysis of unidirectional com-
posites, J. Compos. Mater., 6, 48, 1972.
53. Dvorak, G.J., Rao, M.S.M., and Tarn, J.Q., Yielding in unidirectional composites
under external loads and temperature changes, J. Compos. Mater., 7, 1973.
54. Liu, Y.J. and Chen, X.L., Evaluations of the effective material properties of carbon
nanotube-based composites using a nanoscale representative volume element, Mech.
Mater., 35, 69, 2003.
55. Chen, X.L. and Liu, Y.J., Square representative volume elements for evaluating the
effective material properties of carbon nanotube-based composites, Comp. Mater. Sci.,
29, 1, 2004.
56. Li, C.Y. and Chou, T.W., Multiscale modeling of carbon nanotube reinforced polymer
composites, J. Nanosci. Nanotechnol., 3, 423, 2003.
57. Fisher, F.T., Bradshaw, R.D., and Brinson, L.C., Fiber waviness in nanotube-
reinforced polymer composites I. Modulus predictions using effective nanotube
properties, Compos. Sci. Technol., 63, 1689, 2003.
58. Bradshaw, R.D., Fisher, F.T., and Brinson, L.C., Fiber waviness in nanotube-
reinforced polymer composites II. Modeling via numerical approximation of the
dilute strain concentration tensor, Compos. Sci. Technol., 63, 1705, 2003.
59. Levin, R. N., Quantum Chemistry, Prentice Hall, Englewood Cliffs, NJ, 1999.
60. Mylvaganam, K. and Zhang, L.C., Chemical bonding in polyethylenenanotube com-
posites: a quantum mechanics prediction, J. Phys. Chem. B, 108, 5217, 2004.
61. Jaffe, R.L., Quantum chemistry study of fullerene and carbon nanotube fluorination,
J. Phys. Chem. B, 107, 10378, 2003.
62. Bauschlicher, C.W., High coverages of hydrogen on a (10,10) carbon nanotube, Nano
Lett., 1, 223, 2001.
63. Bauschlicher, C.W. and So, C.R., High coverages of hydrogen on (10,10), (9,0) and
(5,5) carbon nanotubes, Nano Lett., 2, 337, 2002.
64. Bauschlicher, C.W., Hydrogen and fluorine binding to the sidewalls of a (10,10)
carbon nanotube, Chem. Phys. Lett., 322, 237, 2000.
ical vapor deposition of diamond films, Phys. Rev. B, 42, 9458, 1990.
66. Frankland, S.J.V., Harik, V.M., Odegard, G.M., Brenner, D.W., and Gates, T.S., The
stressstrain behavior of polymernanotube composites from molecular dynamics
simulation, Compos. Sci. Technol., 63, 1655, 2003.
67. Griebel, M. and Hamaekers, J., Molecular dynamics simulations of the elastic moduli
of polymer-carbon nanotube composites, Comp. Meth. Appl. Mech. Engin., 193, 1773,
2004.
68. Cornell, W.D., Cieplak, P., Bayly, C.I., Gould, I.R., Merz, K.M., Ferguson, D.M.,
Spellmeyer, D.C., Fox, T., Caldwell, J.W., and Kollman, P.A., A second generation
force field for the simulation of proteins, nucleic acids, and organic molecules, J. Am.
Chem. Soc., 117, 5179, 1995.
69. Allinger, N.L., Yuh, Y.H., and Lii, J.H., Molecular mechanics. The MM3 force field
for hydrocarbons, J. Am. Chem. Soc., 111, 8551, 1989.
70. Duffy, E.M., Kowalczyk, P.J., and Jorgensen, W.L., Do denaturants interact with
aromatic hydrocarbons in water? J. Am. Chem. Soc., 115, 9271, 1993.
71. Smith, J.C. and Karplus, M., Empirical force field study of geometries and con-
formational transitions of some organic molecules, J. Am. Chem. Soc., 114, 801,
1992.
72. Rappe, A.K. and Casewit, C.J., Molecular Mechanics across Chemistry, University
Science Books, Sausalito, CA, 1997.
73. Allen, M.P. and Tildesley, D.J., Computer Simulation of Liquids, Oxford University
Press, Oxford, 1987.
74. Frankland, S.J.V. and Harik, V.M., Analysis of carbon nanotube pull-out from a
polymer matrix, Surf. Sci., 525, L103, 2003.
75. Hu, Y.H. and Sinnott, S.B., Molecular dynamics simulations of polyatomic-ion beam
deposition-induced chemical modification of carbon nanotube/polymer composites,
J. Mater. Chem., 14, 719, 2004.
76. Hu, Y., Jang, I., and Sinnott, S.B., Modification of carbon nanotube-polystyrene
matrix composites through polyatomic-ion beam deposition: predictions from molec-
ular dynamics simulations, Compos. Sci. Technol., 63, 1663, 2003.
77. Werder, T., Walther, J.H., Jaffe, R.L., Halicioglu, T., Noca, F., and Koumoutsakos,
P., Molecular dynamics simulation of contact angles of water droplets in carbon
nanotubes, Nano Lett., 1, 697, 2001.
78. Halicioglu, T. and Jaffe, R.L., Solvent effect on functional groups attached to edges
of carbon nanotubes, Nano Lett., 2, 573, 2002.
79. Werder, T., Walther, J.H., Jaffe, R.L., Halicioglu, T., and Koumoutsakos, P., On the
water-carbon interaction for use in molecular dynamics simulations of graphite and
carbon nanotubes, J. Phys. Chem. B, 107, 1345, 2003.
80. Liang, Z., Gou, J., Zhang, C., Wang, B., and Kramer, L., Investigation of molecular
interactions between (10,10) single-walled nanotube and Epon 862 resin/DETDA
curing agent molecules, Mater. Sci. Engin. A, 365, 228, 2004.
81. Wei, C.Y., Srivastava, D., and Cho, K.J., Thermal expansion and diffusion coefficients
of carbon nanotubepolymer composites, Nano Letters, 2, 647, 2002.
82. Odegard, G.M., Frankland, S.J.V., Herzog, M.N., Gates, T.S., and Fay, C.C., Consti-
tutive modeling of crosslinked nanotube materials, in 45th AIAA/ASME/ASCE/AHS/
ASC Structures, Structural Dynamics, and Materials Conference, Palm Springs, CA,
2004.
83. Odegard, G.M., Gates, T.S., Nicholson, L.M., and Wise, K.E., Equivalent-continuum
modeling of nano-structured materials, Compos. Sci. Technol., 62, 1869, 2002.
84. Fukuda, H. and Takao, Y., Thermoelastic properties of discontinuous fiber composites,
in Chou, T.W., Comprehensive Composite Materials Volume 1: Fiber Reinforcements
and General Theory of Composites, Elsevier, New York, 2000.
85. Brenner, D.W., Shenderova, O.A., Harrison, J.A., Stuart, S.J., Ni, B., and Sinnott,
S.B., Second generation reactive empirical bond order (REBO) potential energy
expression for hydrocarbons, J. Phys. C: Condensed Matter, 14, 783, 2002.
86. Fung, Y.C., Foundations of Solid Mechanics, Prentice-Hall, Englewood Cliffs, NJ,
1965.
87. Nielsen, O.H. and Martin, R.M., Quantum-mechanical theory of stress and force,
Phys. Rev. B, 32, 3780, 1985.
88. Vitek, V. and Egami, T., Atomic level stresses in solids and liquids, Physica Status
Solidi. B: Basic Research, 144, 145, 1987.
24
Introduction to
Nanoscale, Microscale,
and Macroscale Heat
Transport:
Characterization
and Bridging of Space
and Time Scales
Christianne V.D.R. Anderson and Kumar K. Tamma
CONTENTS
24.1 Introduction
24.2 Spatial and Temporal Regimes in Heat Conduction
24.3 Considerations in TimeHeat Conduction
24.4 Considerations in SizeHeat Conduction
24.5 Boltzmann Transport Equation
24.5.1 One-Temperature Models
24.5.2 C- and F-Processes Model A Unied Theory
24.5.3 Validation of Thermodynamics Second Law for
C- and F-Processes Model
24.5.3.1 Fourier Model
24.5.3.2 Cattaneo Model
24.5.3.3 Jeffreys-Type Model
24.6 Two-Temperature Models
24.7 Relaxation Time
24.8 Numerical Illustration Two-Temperature Model and Pulse
Laser Heating
24.9 Numerical Illustration One-Temperature Model and Heat
Conduction Model Number F
T
24.9.1 Spanning Spatial Scales
24.10 Multilayers and Superlattices
24.11 Equation of Phonon Radiative Transfer (EPRT)
24.12 CallawayHollands Model
24.12.1 Comparison of Holland Model and EPRT Results
24.13 Molecular Dynamics
24.13.1 Pristine Nanotube
Problems
Acknowledgments
References
24.1 INTRODUCTION
This chapter serves as an introduction and a prelude to understanding the fundamentals
encompassing nanoscale, microscale, and macroscale heat transport. Because of the
continued trend in the miniaturization of devices and rapid progress in the design and
synthesis of materials and structures leading to the notion of materials by design, there
is signicant interest in improved understanding of heat transport across the broad
spectrum of length scales ranging from nanoscale to macroscale regimes and time
scales encompassing nite to innite speeds of heat propagation.
Since the early work on size effects on thermal conductivity in 1938 by Haas
and Biermasz and theoretical work by Casimir,
1
and more recent efforts demonstrating
that the thermal conductivity of thin lm structures is one to two orders of magnitude
smaller than its bulk counterpart,
2
from an engineering perspective many researchers
have shown a keen interest in providing an improved scientic understanding of
thermal transport at small length scales such as the nanoscale regime. Some early
works in the 1950s and 1960s by Klemens,
3
Callaway,
4
and Holland
5
emanating from
the Boltzmann transport equation (BTE) have indeed provided viable approaches to
modeling lattice thermal conductivity based on phonon-scattering processes. Alter-
nately, since the early 1990s, developments emanating from the BTE with relation to
the equation of radiation transfer (ERT), namely the equations for phonon radiative
transfer (EPRT)
6,7
and others such as Monte Carlo
811
and molecular dynamics (MD)
simulations,
1216
have also received considerable research activity. While MD simula-
tions do not require any tting parameters, they lack the ability to perform large-
scale simulations at desired length scales that are of practical interest. Alternately,
the BTE-based approaches invariably need tting parameters(for example, the relax-
ation time, the acoustic phonon dispersion from property of bulk medium, and the
like). On the other hand, these more recent efforts since the 1990s attempt to bring
such physics into an engineering perspective and follow the same methods as in
Callaway
4
and Holland
5
or improvements thereof and stem serious concerns in
employing phenomenological models such as the classical Fourier model and other
modications such as that of CattaneoVernotte.
17,18
Consequently, these more recent
efforts have led to microscopic energy transfer models, and although they also need
to employ tting parameters, they appear to show reasonably good t to experimental
data.
6,9,19,20
The primary reasons and arguments for directing research efforts to
microscopic energy transfer models and other formulations that have been cited in the
literature since the 1990s, for example, in the prediction of thermal conductivity of very
thin lms, note that (1) the Fourier law overpredicts the actual heat ux,
6
resulting in
thermal conductivity results that could be as much as two orders of magnitude greater
than experimental reported values for thin dielectric lms; (2) the Cattaneo equation
lacks the ability to capture the ballistic transport process;
8,19
and (3) for processes that
are small in both space and time, the energy balance fails based on the phenomenological
Fourier model.
9
Taking some of these arguments at face value, a reference is also made
similar to issue (1) by the present authors;
21
however, in this chapter we accurately put
into context arguments related to all the preceding issues. This chapter fundamentally
places into context and provides challenging answers to the aforementioned arguments
existing in the literature, and the reader is encouraged to analyze the underlying details
and draw relevant conclusions to understanding thermal transport across a broad range
of spatial scales and also temporal scales ranging from nite to innite speeds of heat
propagation.
24.2 SPATIAL AND TEMPORAL REGIMES
IN HEAT CONDUCTION
Quantum mechanics help us recognize that solids are composed of several discrete
particles (electrons, neutrons, and protons) at the microscopic level. Depending on
the type of solid, different mechanisms of heat transport take place in materials in
different manners and are called heat carriers. Modeling the spatial and temporal
regimes of materials can be broadly subdivided into four categories
22
:
Macroscale
Mesoscale
Microscale
Nanoscale
The characteristic lengths and times for these regimes are depicted in Table 24.1.
In heat conduction, two major areas have been prominent: (1) the macroscale, which
involves the modeling of macroscopic effects; and (2) the microscale, which encom-
passes all the other regimes in materials modeling (nanoscale effects are implied).
TABLE 24.1
Characteristic Lengths and Times in Materials Modeling
Regime Characteristic Length Characteristic Time
Macroscale 10
3
m 10
3
s
Mesoscale ~10
4
10
7
m ~10
3
10
9
s
Microscale ~10
6
10
8
m ~10
8
10
11
s
Nanoscale ~10
7
10
9
m ~10
10
10
14
s
Source: Zhigilei, L.V., MSE 524: Modeling in Materials Science, Spring 2002.
The macroscale formulation in heat conduction is based on the continuum assump-
tion. It does not consider the size and time dependence of the heat transport. Heat is
assumed to be carried by the atoms. The microscale formulation in heat conduction
considers the physical mechanisms of heat transport through heat carriers where size
and time dependence are crucial. This leads to the explanation of several parameters
that characterize the microscopic regimes in both temporal and spatial regimes.
24.3 CONSIDERATIONS IN TIMEHEAT
CONDUCTION
Besides the notion of temporal scale aspects ranging from the nite nature of heat
propagation to innite speeds of thermal transport, the important time parameters
that govern the microscopic heat transport are as follows
9
:
The thermalization time The time for the electrons and lattice to reach
equilibrium
The diffusion time The time the heat information takes to travel through
the specimen
The relaxation time The time associated with the speed at which a
thermal disturbance moves through the specimen
The heating time The time that an external source heats the specimen
The physical process time The total time duration of interest
When the physical process time is comparable to any of the aforementioned times,
the time effect becomes important and the process is considered to be microscopic
in time. When the heating time is on the order of the thermalization time, the energy
deposition must be considered. The immediate question that naturally arises from
this consideration is whether the energy is deposited in the lattice, the electron, or
both. Finally, when the heating time is comparable to the diffusion time or the
relaxation time, a nite speed of the thermal propagation must be considered.
24.4 CONSIDERATIONS IN SIZEHEAT CONDUCTION
The important size parameters that govern the heat transport in a device are as
follows
9
:
The mean free path ()
The characteristic dimension of the material (L)
From a physical viewpoint, when the mean free path is much less than the charac-
teristic dimension of the material ( << L), the heat transport is said to be macroscopic
and is commonly termed the purely diffusive limit. In this limit, there exist enough
scattering mechanisms within the lm. In dielectrics, these scattering mechanisms
help bring the phonons within the lm back to equilibrium and help establish a
temperature gradient.
When the mean free path is on the order of or much greater than the characteristic
dimension of the material ( L or >> L), the heat transport is said to be microscopic,
and the transport presents itself in a partially diffusive ballistic and purely ballistic
manner, respectively. As the size of the material is decreased (and in comparison to
becomes small), so do the scattering mechanisms within the lm, and a temperature
gradient might not be established at the ballistic limit. As observed at the purely ballistic
limit, there exist temperature jumps at the physical boundaries (analogous to slip
conditions), and the nal temperature prole is characterized by .
23
Since a temperature gradient is not established at the purely ballistic or partially
ballistic diffusive limits, then the Fourier law in its purest denition breaks down, thus
making it impossible to predict the thermal conductivity.
These limits can be observed in Figure 24.1. The size effects to thermal con-
ductivity were originally observed in 1938 in the experimental works of Haas and
Biermasz, and the theoretical explanation soon followed in the works of Casimir.
1
Casimir noted that at the purely ballistic regime a temperature gradient could not
be established, and strictly according to the Fourier law, it was not possible to
prescribe the thermal conductivity. At the purely ballistic limit, the temperature at
the boundaries, and not the temperature gradient within the lm, governs the heat
transport. Since thermodynamic equilibrium is restored due to the scattering of
particles from the boundaries, the heat conduction by these particles is suggested to
be similar to conduction by photons and can be analyzed as a radiative transfer model
where the heat ux across the specimen can be described by
(24.1)
where is the StefanBoltzmann constant for the heat carrier, and T
1
and T
2
are the
temperatures of the faces of a thin lm. This is commonly referred to as the Casimir limit.
At this juncture, some noteworthy questions naturally arise. What are the under-
lying issues and approaches for characterization and bridging space and time scales?
When do we need to resort to nonclassical approaches and what are the limits of
FIGURE 24.1 Comparison of heat transport characterization.
L >> , diffusive
transport
>> L, ballistic
transport
~ L, ballistic-
diffusive transport
T
1
T
2
T
2
T
2
T
1
T
1
L L
L
Fourier Casimir limit What is the model that
characterizes this limit?
x
x
x
[( )/ ]
/
T T
L R
+
4 4
2
1 4
q T T
( )
1
4
2
4
the classical models? Is there a unied theory underlying heat transport that best
describes the conduction of heat spanning the various scales, and what are the
limitations? From a modeling and simulation viewpoint, what are the simplest but
effective formulations that are easy to implement and are not computationally inten-
sive? Several of these issues are addressed in the text to follow and the various
methods are also briey highlighted.
24.5 BOLTZMANN TRANSPORT EQUATION
The heart of heat transport theory is the Boltzmann transport equation (BTE). The
BTE determines the status of a particle via its location and velocity. The most general
form of the BTE
24
is given as
(24.2)
where is the nonequilibrium thermodynamic distribution function, v()
is the phonon velocity (that is, the speed of sound in dielectrics), a(v) is the particle
acceleration, and t is time. The rst term in Equation 24.2 represents the net rate of
particles over time, the second term is the convective inow of particles in physical
space, the third term is the net convective inow due to acceleration in velocity
space, and the term on the right-hand side is the net rate of change of particles inside
a control volume due to collisions. The phonon velocity in a dielectric material is
fairly constant over a large frequency range, and thus f/v can be neglected. Both
energy and temperature gradients tend to disturb the electron distribution, and this
tendency is opposed by processes that restore equilibrium, such as the scattering of
electrons or phonons by lattice vibrations and crystal defects. As a result, the
scattering term is approximated under the relaxation time approximation
25
as
(24.3)
where is the thermodynamic distribution at equilibrium (BoseEinstein
distribution for boson particles [such as phonons] and FermiDirac distribution for
fermion particles [such as electrons]), f
0
/t 0, and (, v) is the rate of return
to equilibrium and is called the relaxation time.
Kinetic theory is derived from the BTE under the premise that there is local
thermal dynamic equilibrium(LTE).
26
In cases where LTE is not achieved, one solves
the BTE. However, under the presence of a temperature gradient, the LTE is implied,
and for simplicity in illustrating the basic concepts we consider the one-dimensional
form of the BTE. The diffusion term in Equation 24.2, f/x f
0
/x, and the f /x
term can be approximated as
4
(24.4)
+ +

_
,
f
t
v f a
f
v
f
t
scatt
f r T ( , , )
_
,

f
t
f f
t
f f
scatt
( )
0 0
f r T
0
( , , )

f
x
df
dT
dT
dx
0
The ux of particles is given by
25
(24.5)
The transient one-dimensional BTE under the relaxation time approximation and
the temperature gradient approximation is given by
(24.6)
As the temperature of a material is raised or decreased, these particles interact and
collide with each other in order to return the system to thermodynamic equilibrium (the
mean time associated with these scattering processes is the relaxation time [], and the
mean distance associated with these scattering processes is the mean free path []).
24.5.1 ONE-TEMPERATURE MODELS
Consider the MaxwellBoltzmann, BoseEinstein, and FermiDirac distribution func-
tions shown in Fig. 24.2 where the Fermi energy is set to zero. At high frequencies
(or energies), the effect of the FermiDirac and the BoseEinstein statistical functions
are eliminated and all distributions converge to a MaxwellBoltzmann distribution.
MaxwellBoltzmann distribution function:
(24.7)
FIGURE 24.2 Occupancy probability for the FermiDirac, BoseEinstein and MaxwellBolt-
zmann distributions. (From Van Zghbroeck, B.J., Distribution Functions, 1996.)
200%
150%
100%
50%
0%
0.2 0.1 0 0.1 0.2
Energy (eV)
O
c
c
u
p
a
n
c
y

p
r
o
b
a
b
i
l
i
t
y
q x v f x D d
x
D
( ) ( ) ( )
+
f
t
v
df
dT
dT
dx
f f x
x
0 0
( )
f E f E
Ae
MB
E
T
B
( ( )) ( ( )) :
( )

1
BoseEinstein distribution function:
(24.8)
FermiDirac distribution function:
(24.9)
where A is a normalized constant,
B
is the Boltzmanns constant, and E
F
is the
Fermi energy. If the frequency is large enough, in the foregoing distribution
functions we have
(24.10)
Therefore, we have
(24.11)
At high frequencies, the wavelength of the particle is short and the statistical
distinction is unimportant. Multiplying Equation 24.6 by and integrating
over the two separate frequency ranges [0,
T
] and [
T
,
D
], based on the physics
of the actual distribution functions, because of this statistical unimportance, it is
reasonable to assume that , due to the observation that
the distribution function is fairly constant over time for high frequencies. In other
words, as the different distribution functions reach their equilibrium positions, most
of the change in the distribution functions occurs at the lower frequencies, and the
distributions at the high frequency tail are fairly constant over time. Hence, the
df/dt term is neglected. Also note that as time evolves, the threshold frequency must
change as the distributions reach these equilibrium stages. This is a similar analogy
to the change of the Fermi energy with temperature (see Reference 27).
Using this fundamental basis, we now highlight a novel C- and F-processes
model based on the hypothesis that upon the application of a temperature gradient,
there simultaneously coexist both slow processes (at low energies and termed C-
processes) and fast processes (at high energies and termed F-processes) associated
with the heat carriers and evolving with time as the processes proceed from a nite
nature of propagation to a process involving innite speeds.
1719
The spanning of
space scales is in the sense of characterizing ballistic to diffusive limits, and the
spanning of the time scales is in the sense of characterizing nite to innite speeds
of heat propagation. As with BTE-based approaches, certain limitations exist; none-
theless, coupling of the simple yet effective C- and F-processes model with compu-
tationally attractive variants of MD or lattice dynamics to avoid the need for tting
parameters may provide extensions to circumvent some of the limitations and pro-
vide extensions for multiscale capabilities.
f E f E
Ae
BE
E
T
B
( ( )) ( ( )) :
( )

1
1
f E f E
Ae
FD
E E
T
F
B
( ( )) ( ( )) :
( ( ) )

1
1
f E
T D
( ( )) , [ , ] 0

f E
t
T D
( ( ))
, [ , ]
0

x
D ( )
v D d
x
f
t
T
D

( )

equals 0
24.5.2 C- AND F-PROCESSES MODEL A UNIFIED THEORY
Following the previous discussion, we now dene the total heat ux as
(24.12)
where the postulation is that the integral up to a threshold frequency
T
involves
the slow C-processes and is termed q
C
, and that the integral from the threshold to
innity involves the fast F-processes and is termed q
F
.
As discussed previously, considering the total heat conduction process, the
BTE can be multiplied by and integrated over the entire frequency
range. Also, note that (not detailed here) due to the obser-
vation that the distribution function is constant over time for high frequency;
therefore, df/dt 0.
Next, introducing the nondimensional heat conduction model number F
T
as the
following ratio, we have
(24.13)
This is physically depicted in Reference 28 and has the interpretation
(24.14)
Further, the following is dened
(24.15)
Upon application of the heat ux of particles given in Equation 24.12 and the thermal
conductivity given in Equation 24.15, with the denition from Equation 24.13, the
transient BTE nally yields the C- and F-processes heat conduction constitutive
model in terms of the heat conduction model number as
(24.16)
(24.17)
(24.18)
q v f D d v f D d q q
x x C F
T
D T
+ +

( ) ( )
0
v D
x
( )
v D d
x
f
t T
D

( )

0
F
v D d
v D
T
x
df
dT
x
df
dT
T
D
D
2
2
0
0
0
( )
( ))d
F
K
K K
T
F
F C
K
F
Conductivity ( ) associated with f fast F-Processes

Conductivity ( ) associated K
F
with fast F-processes + Conductivity( ) ass K
C
oociated with slow C-processes
K v
df
dT
D d v
df
dT
D
x x
T
T
D
+

2
0
0
2
0

( ) ( ))d K K
C F
+
q q q
F C
+
q F K
dT
dx
F T

q
dq
dt
F K
dT
dx
C
C
T
+ ( ) 1
In general, the C- and F-processes heat conduction constitutive model is given by
(24.19)
(24.20)
(24.21)
which explains the present derivation based on fundamental physical principles
emanating from the Boltzmann transport equation.
In the previous equations, q is the total heat ux due to the mechanisms of heat
conduction (composed of those associated with each of the Fourier-type fast pro-
cesses and the Cattaneo-type slow processes), and K is the total conductivity, which
is the sum of the Fourier (effective) conductivity, K
F
, and the Cattaneo (elastic)
conductivity, K
C
. Thus, K K
F
+ K
C
.
When F
T
1, the right-hand side of Equation 24.20 is zero and the total heat
ux is given by the Fourier law. This implies that the conductivity associated with
the slow C-processes K
C
0, or in other words, most of the dominant transport with
evolution of time to steady state is via the fast F-processes. On the other hand, when
F
T
0, the right-hand side of Equation 24.19 is zero and the total heat ux is given
by the Cattaneo law. Consequently, K
F
0 and most of the dominant transport with
the evolution of time to steady state is via the slow C-processes. Thus, in the limiting
cases of F
T
the C- and F-processes model can recover both Cattaneo and Fourier laws.
The combined representation of the C- and F-processes model can be shown by
adding Equations 24.17 and 24.18 to yield
(24.22)
By substituting the C- and F-processes postulation that the total heat ux is a com-
bination of the ux associated with the low-energy processes and the high-energy
processes (see Equation 24.12) where q
C
q q
F
, and the high-energy processes
are described by a Fourier-like heat ux given by q
F
K
F
dT/dx, Equation 24.22
reduces to
(24.23)
Rearranging Equation 24.23, we obtain the so-called Jeffreys model of heat
conduction
29
as
(24.24)
q
F T
F K T
q
q
C
C
T
t
F K T +

( ) 1
q q q +
F C
q
q
+
d
dt
K T
C
q
q
+

d K T
dt
K T
F
( ( ))
q
q
+ +
1
]
1

d
dt
K T
K
K
d
dt
T
F
( )
where the ux q, the thermal conductivity K, and the relaxation time are the total
contributions, and K
F
/K is dened as the retardation time
R
.
29
By adhering to the
notion of the heat conduction model number, it is nally possible to characterize
the energy transport of the heat conduction process from transient to steady state.
It is noteworthy to mention that the Jeffreys model also reduces to the Cattaneo
model when K
F
0. However, when K
F
K it only reduces to a Fourier-like model
that contains the elusive relaxation time parameter , unlike the C- and F-processes
model that identically yields the Fourier model.
By substituting Equations 24.19 through 24.21 into the energy equation, we elim-
inate the ux and obtain the generalized one-step (GOS) C- and F-processes one-
temperature formulation as
(24.25)
(24.26)
where T is the total temperature, K is the total thermal conductivity given by kinetic
theory as K C
2
/3, the temperature propagation speed is given by ,
the thermal diffusivity is given by K/c, C is the specic heat of the heat carrier,
is the density of the heat carrier, S is the external heat source, and the retardation
time is given by
R
F
T
.
In conjunction with the C- and F-processes model , we argue following the initial
work by Klitsner and colleagues
23
on phonon radiative heat transfer where heat
conduction by phonons can be analyzed as radiative transfer that the phonons are
emitted from the surface and that the boundary conditions are developed based on
an energy balance at the surface as q
n
q
rad
, where
(24.27)
where T represents the surface temperatures of the left and right sides of the lm,
and T
is the ambient temperature on the right and left of the lm. It is assumed
that the thermalizing black boundaries (the ambients) are the boundaries of the metal
lm between which the dielectric lm is enclosed.
Finally, after the correct heat ux is computed via the CF model (which must
be a constant across the lm due to the radiative equilibrium assumption at steady
state), an effective thermal conductivity is obtained. We invoke the following prop-
osition that the thermal conductivity provided by the CF model is due to the heat
ux within the lm based on the established boundary temperatures divided by the
temperature gradient based on the imposed ambient temperatures. Thus,
(24.28)
L T f T T T T K c S S F
xx T T 1 1
( ) ( , , , , , , , , , , ),

for FF
T
< 1
L T f T T T K S F
xx T 2 2
1 ( ) ( , , , , , ),
for
c K c
T
/
q q n f C v T T x L
n

( , , , ), , for 0
K
C F

true heat flux within the film due to established boundary temperatures
temperature gradient due to the imposed ambient temperatures
24.5.3 VALIDATION OF THERMODYNAMICS SECOND LAW
FOR C- AND F-PROCESSES MODEL
The thermodynamics second law is represented in terms of the ClausiusDuhem
inequality as
(24.29)
where s is the entropy, and r is the heat source.
It can be shown that for a solid,
(24.30)
The thermodynamics second law can also be represented in the form
(24.31)
Therefore, the thermodynamics second law can be simply decoupled to yield the
following two sufcient conditions:
(24.32)
(24.33)
where the rst equation represents no viscous dissipation of kinetic energy into
thermal energy, and the second condition represents that in heat conduction, the heat
ow is from a high-temperature region to a low-temperature region. From the
previous section, we showed that the processes in the C- and F-processes heat
conduction model belong to two different frequency ranges. Therefore, the F-pro-
cesses and the C-processes are independent, and either of the processes indeed
satises the thermodynamics second law independently, namely,
(24.34)
(24.35)
To satisfy the thermodynamics second law, the heat conduction model number F
T
has to strictly satisfy the bounds F
T
[0,1]. For other cases such as F
T
< 0 or F
T
>
1, the thermodynamics second law will be violated. To be consistent with the second
law of thermodynamics, one must ensure 0 k
F
and 0 k
C
; hence, 0 k k
F
+ k
C
.
Thus, from its basic underlying denition, the bounds are strictly F
T
[0, 1]. For the
Ds
Dt
rT T
_
,

1 1
( ) q
T
Ds
Dt
r
_
,
+ q 0
( ) Ts r T T +
q q
1
0
( ) Ts r + q 0
q T 0
q
F T
T F k T ( )
2
0
q
C
T
t s
t
T
F k
T e T s ds

( )
( , )
1
0

x
case of F
T
< 0, it leads to 0 < k
F
< k
C
or 0 < k
C
< k
F
, and for the case of 1 < F
T
,
it leads to 0 < k
F
< k
C
or 0 < k
C
< k
F
. For either of the previous two underlying
cases, the second law of thermodynamics is violated, and hence it is neither valid
nor physically acceptable. Some existing classical models highlighted next can all
be readily explained via the C- and F-processes model as described previously.
24.5.3.1 Fourier Model
Under the assumption of a macroscopic continuum formulation, this phenomeno-
logical model proposed by Fourier
30
is given by
(24.36)
where k is the conductivity tensor, q is the heat ux, and T is temperature, which
together with the energy equation yields the parabolic one-step (POS) heat conduc-
tion equation or the parabolic heat conduction (PHC) equation and is diffusive with
the notion of innite speed of propagation of thermal disturbances.
24.5.3.2 Cattaneo Model
Because of the anomalies associated with the Fourier model and in order to account
for a nite temperature propagation speed, the Cattaneo-type model
17,18
is based on
the notion of relaxing the heat ux and is given as
(24.37)
where is the relaxation time (Cattaneo originally developed this model for gases).
For the Cattaneo model, the temperature propagation speed is given as
(24.38)
where is the material density, and c is the heat capacity. Together with the energy
equation, it yields the hyperbolic one-step (HOS) heat conduction equation or the
hyperbolic heat conduction (HHC), which is propagative with the notion of a nite
speed of heat propagation.
24.5.3.3 Jeffreys-Type Model
The original Jeffreys model
31,32
was originally proposed for studying the wave
propagation in the earths mantle,
33
and is given as
(24.39)
q k T

q
q
t
k T
c
k
c
T

+

_
,
1 0 2
t t

which relates the stress and the rate of strain tensor and
0
is the zero shear
rate viscosity. In this constitutive model originally proposed in 1929,
1
is dened
as the relaxation time and
2
is dened as the retardation time; no mention was
ever made by Jeffreys relating this to the notion of heat. Based on experiences with
viscoelastic uids, a Jeffreys-type heat ux phenomenological model drawn from
the original Jeffreys equation (Equation 24.39) was subsequently proposed and
introduced by Joseph and Preziosi
29
for heat conduction, which appears as
(24.40)
where is the relaxation time and K k
1
/k is the retardation time. It is important
to note that in the original model proposed by Jeffreys,
33
which relates stress and
strain rate, or in the corresponding heat ux model introduced by Joseph and
Preziosi,
29
which relates the heat ux to the temperature gradient, the retardation
time can never exceed the relaxation time because it makes the physical interpretation
meaningless if violated. When selecting the retardation time, K 0, the Jeffreys-
type model degenerates to the Cattaneo model as
(24.41)
and when selecting the retardation time equal to the relaxation time, K , the Jeffreys-
type model only degenerates to a Fourier-like diffusive model with relaxation as
(24.42)
24.6 TWO-TEMPERATURE MODELS
Likewise, there also exists the so-called microscale two-temperature theory (namely,
the two-step temperature equations describing nonequilibrium behavior between
electrons and phonons). From a microscale viewpoint, the pioneering work of the
two-temperature diffusive formulation of Anisimov et al.,
34
who seem to have devel-
oped the rst parabolic two-step (PTS) energy balance equations, is signicant and
given as
(24.43)
(24.44)
(24.45)
( , ),
q
q
+

1
]
1
t
k T K
T
t
( )

q
q
t
k T
q
q
+

1
]
1

t
k T
T
t
( )
C T
T
t
G T T S t
e e
e e
e l
( ) ( ) ( , )
+
q
r
r
C
T
t
G T T
l
l
e l
( )
q
e e e
k T
where S(r,t) is the internal heating due to a laser pulse and is an integral part of the
two-temperature theory, C
e
is the electron heat capacity and C
l
is lattice heat capacity
(with C C
e
+ C
l
), G is the electron-lattice coupling factor whose experimental
determination has been challenged,
35,36
and k
e
is associated with the electron thermal
conductivity in steady state. The electron heat ux q
e
pertains to the Fourier model,
and diffusion in the lattice is neglected during the transient duration.
The generalized two-step (GTS) pulse heating model for metals recently proposed
by Tamma and Zhou
37
allows not only choices of propagation or diffusion transport
mechanisms at the microscale levels but also transitions via the simultaneous intro-
duction of microscale relaxation and retardation times.
We next introduce the physical notion of a heat conduction model number, F
T
e
,
associated with the electron heat ux involving F-processes and C-processes as
(24.46)
For microscale heat transport it now has a different physical meaning than in the
macroscale formulation, although it has similarities with the associated macroscale
formulation and is physically dened in terms of a heat conduction model number,
F
T
e
[0,1], as
(24.47)
where k
eF
and k
eC
are contributions of the F-processes and C-processes to the total
electron thermal conductivity given by k
e
k
eF
+ k
eC
. Furthermore, the aforemen-
tioned constitutive formulation for the two-step theory also has a different physical
meaning than that of the macroscale formulation in the following sense.
We consider for microscale heat transport
(24.48)
where q
e
is the total electron heat ux and is composed of the sum of the electron
heat ux associated with the F-processes, q
eF
, and the C-processes, q
eC
, respectively.
Employing this constitutive model into the Anisimov et al.
34
two-step energy
equations yields the generalized microscale electron-lattice two-step model equations
q
e e e
e
eF e
e e
e
k T
t
k
t
T
k T
k

q
q
( )
1
]
1
t
F
t
T
T e
e
( )
F
k
k k
T
eF
eF eC
e

+
q
q
q
eF eF e T e e
eC
eC
eC e
k T F k T
t
k T
e

+

1 ( )
+
F k T
T e e
e eF eC
e
q q q
with F
T
e
[0, 1] of which the GTS pertains to the range F
T
e
(0,1), which is parabolic
(diffusive) in nature:
(24.49)
The preceding GTS model is parabolic (diffusive), and the transmission of informa-
tion is immediately felt everywhere. It is also to be noted that by approximate
mathematical manipulations, the aforementioned equations may be combined to lead
to a single equation for the determination of the electron or lattice temperatures,
respectively.
24.7 RELAXATION TIME
The rst step in solving for the heat transport at the different length scales is to
approximate the elusive relaxation time, . Accounting for all the scattering
mechanisms within a lm can approximate the relaxation time. Accurately quanti-
fying and qualifying all the possible scattering mechanisms proves to be the most
difcult task in solving the CF model, or any model derived from BTE for that
matter.
36,9
This task can be simplied by accounting for all the internal scattering
processes within a specimen by assuming that all scattering mechanisms are
independent. And, using Matthiessens rule, which inversely adds all scattering
contributions as 1/
i

j
1/
i, j
where
i,j
s are the contributions of the various
scattering mechanisms based on i, the modes of polarizations of the phonons, and
j represents the different types of scattering rates. These scattering mechanisms
contribute to the total resistance to the heat transport.
Many researchers
1,5,3842
describe all the different scattering mechanisms that
have been used for predicting relaxation times for doped and undoped single and
polycrystalline dielectric lms. In a nutshell, most of the scattering is due to crystal
imperfections or interactions with other phonons, and at low temperatures (or for
lms where the L) the boundary of a crystal. For the results shown here, we
assume that these are the main scattering mechanisms
(24.50)
where the scattering due to defects is given as
(24.51)
C T
T
t
G T T S
C
T
t
G T T
e e
e e
e l
l
l
e
( ) ( )
(

q
r
ll
e e e
e
e
T e
k T
k t
F
t
T
e
)
( ) q
q
+

1
]
1

1 1 1 1

+ +
defect U GB

defect

1
c
v
where is a constant (usually one),
c
is the scattering cross-section, is the level
of impurity in the medium (in other words, the number of scattering sites in the
medium), and v is the speed of the phonons, which for dielectrics is the speed of
sound. The scattering cross-section is approximated as
6
(24.52)
where R is the radius of the lattice imperfection, s is the size parameter given by
s 2R/
dominant
, and
dominant
is the dominant wavelength of phonons denoted by
hv ~
dominant
K
B
T, where K
B
is the Boltzmann constant, and h is Plancks constant.
The scattering due to other phonons is accounted for by the U-process scattering as
6
(24.53)
where A is a nondimensional constant that depends on the atomic mass, the lattice
spacing, and the Grinessen constant;
D
is the Debye temperature; is the fre-
quency; T is the temperature; and a is a parameter representing the effect of the
crystal structure.
Finally, the grain boundary scattering is given as
40
(24.54)
where v
s
is the average phonon velocity, and d
g
is the sample size in the direction
perpendicular to heat ow.
24.8 NUMERICAL ILLUSTRATION TWO-TEMPERATURE
MODEL AND PULSE LASER HEATING
For the experiment by Brorson et al.,
43
C
e
1.84E + 4(J/m
3
K), C
l
2.5E + 6(J/m
3
K),
15.3 nm, G 2.6E + 16(W/m
3
K), J 10(J/m
2
), k
eo
310(W/mK), L 0.1 m,
and R 0.093 were employed. Also, t
p
96 fs was used except when the source
was that of case (1) as described subsequently; t
p
100 fs was employed for the
experimental simulations.
For the selected experiment from Brorson et al.
43
for short-pulse laser heating
of thin gold lms, the following were investigated:
Case 1: For the source, the actual measured autocorrelation of the laser pulse
was rst employed.
44
Case 2: The source was specically selected as
(24.55)

c
R
s
s
_
,
2
4
4
1

U
D
aT
A
T
e
D
GB
g
s
d
v
S
R
t
J e
p
x t
t
p
_
,
_
,
0 94
1
2 77
2
.
.

where t
p
is the full width at half maximum of the pulse duration and is the radiation
penetration depth.
Furthermore, for the relaxation parameter, , the following were investigated:
(1) accounts for the temperature dependence with data taken from Pells and Shiga,
45
and (2) 0.04 ps (constant value). In all the situations, temperature dependence
of the electron thermal conductivity and heat capacity was assumed.
Employing the GTS, studies were conducted to demonstrate the theoretical
consequences alluded to earlier, including efforts to shed further light on the char-
acterization of the heat transport behavior in these experiments. Figure 24.3 shows
the comparative results of the Brorson et al.
43
experiments for the the front surface
with the GTS (F
T
e
0 and permitting [T
l
]; F
T
e
1, which implies 0; and for
illustration, F
T
e
0.5 with [T
l
]). The source was selected from the actual
measured autocorrelation of the laser probe (Case 1). Although not clearly distin-
guishable, the cases of F
T
e
1 and F
T
e
0.5 (arbitrarily selected) seem to be somewhat
closer than F
T
e
0 at very early times in the transient.
Also based on the theoretical premise that there simultaneously exist fast and
slow transport processes, the added objective was to characterize (admittedly, a rst
and primitive attempt) the microscale heat conduction model number, namely the
F
T
e
value, and select this to match closely to the experiment. For example, if F
T
e

0.1 was a close match, then k
eF
0.1k
e
0.1(k
eF
+ k
eC
), which implies that plausibly
FIGURE 24.3 Comparative results for experiment of Brorson et al. (T) and heat source.
(From Bronson, S.D. et al., Phys. Rev. Lett., 59, 1962, 1987.)
1 2 3
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Time (ps)
N
o
r
m
a
l
i
z
e
d

t
e
m
p
e
r
a
t
u
r
e
GTS (Ft = 0)
GTS (Ft = 0.5)
GTS (Ft = 1.0)
Brorson et al. (1987)
90% pertains to transport associated with k
eC
and 10% pertains to transport associated
with k
eF
. Since total thermal conductivity is customarily measured experimentally,
such a plausible challenge was undertaken. However, based on very limited exper-
iments conducted to date, no bounds on errors are available (and the possibility
exists for experimental errors), only a quantitative check could be made and no
conclusive results or inferences could be drawn.
Figure 24.4 shows the comparative results with the source selected instead of
Case 2. The results are in reasonable agreement in the initial part of the transient;
however, some in-depth studies drawing comparisons with the measured autocorre-
lation of the laser pulse need to be conducted.
24.9 NUMERICAL ILLUSTRATION ONE-TEMPERATURE
MODEL AND HEAT CONDUCTION MODEL
NUMBER F
T
Consider the following experiment. Since little literature on experimental results
describing the transient behavior at the ballistic limit exists to our knowledge, we next
present a systematic analysis of the signicance of F
T
spanning the temporal ballistic
limits for the hypothetical case of a diamond lm of 1 m. Considering the lm
structure in Fig. 24.5(a), assume that the curve shown in Fig. 24.5(b) represents
FIGURE 24.4 Comparative results for experiment of Brorson et al. (T) and heat source,
case (2) (From Bronson et al., Phys. Rev. Lett., 59, 1962, 1987.)
1 2 3
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Time (ps)
N
o
r
m
a
l
i
z
e
d

t
e
m
p
e
r
a
t
u
r
e
GTS (Ft = 0.0)
GTS (Ft = 0.5)
GTS (Ft = 1.0)
Brorson et al. (1987)
the characteristic thermal conductivity data at room temperature for the diamond
lm spanning the ballistic to diffusive limits (asymptotically approaching the bulk
data provided by Reference 46 as shown). For the 1-m diamond lm, suppose that
the total thermal conductivity K ~ 2100 W/m/K. Note that the steady-state prediction
is independent of F
T
. We assume that the correct transient behavior can be charac-
terized by F
T
in the CF model where these responses are depicted in Fig. 24.5(c)
with evolution of time to steady state. In the case of the 1-m diamond lm, assume
that F
T
0.5 will correctly characterize the transient to steady-state temperature
FIGURE 24.5 One temperature model and heat conduction model F
T
.
Adiabatic
Adiabatic
T
L
T
R
T
R
T
L
q
.
n
q
.
n
(a) Single film setup
0
500
1000
1500
2000
2500
3000
3500
1e 09 1e 08 1e 07 1e 06 1e 05 0.0001 0.001
T
h
e
r
m
a
l

c
o
n
d
u
c
t
i
v
i
t
y

(
W
/
m
/
K
)
Film thickness (m)
<< L >> L
Film
Bulk
~ L
(b) Thermal conductivity
response of the experiment across the lm thickness. The information that this
hypothetical experiment has provided is that the actual transport behavior for this
lm that is in the partially diffusive ballistic region (due to the observed temperature
jump on boundaries at steady state) also provides a transient response that is 50%
wave-like at a nite speed and 50% diffusive at an innite speed of heat propagation.
Since F
T
0.5, K
F
0.5(K). That is, in obtaining the total conductivity, the conduc-
tivity associated with the fast F-processes is half the total conductivity (the other
half is due to the conductivity associated with the slow C-processes). Based on the
denition of F
T
, for the rst time it is now possible to characterize the elastic
component of the thermal conductivity
29
and provide a fundamental understanding
of the role of the fast F-processes and slow C-processes in the propagation of heat
transport leading to the thermal conductivity property obtained for a given lm.
24.9.1 SPANNING SPATIAL SCALES
At steady state, the nondimensional temperature proles (under the boundary
conditions in Equation 24.27, which are of the third kind) of type IIa diamond
lms spanning lengths from 0.001 to 1000 m based on the CF model imple-
mentation are shown in Fig. 24.6. At room temperature, the mean free path of
diamond lms for experiments performed by Anthony et al.
46
is ~0.447 m (see
Table 24.2). At these lm thicknesses, the temperature plots spanned both the
ballistic ( >> L) and diffusive ( << L) limits. At steady state, all temperature
prole results for various values of F
T
[0,1] were equal as shown in Fig. 24.6.
FIGURE 24.5 (Continued )
300
299.98
299.96
299.94
299.92
299.9
0.0E+00 2.0E 07 4.0E 07 6.0E 07 8.0E 07 1.0E 06
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Film thickness (m)
t = 1
t = 10
t = ss
(c) Transient temperature profile
FIGURE 24.6 Steady-state results for several diamond lm thicknesses using a CF model
with boundary conditions of the third kind. The nondimensional thickness is l/L where l is
the x-axis position and L is the lm thickness.
299.9
299.91
299.92
299.93
299.94
299.95
299.96
299.97
299.98
299.99
300
0 0.2 0.4 0.6 0.8 1
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Nondimensional film thickness
(a) Temperature problems
L = 0.01 m
L = 0.1 m
L = 10 m
L = 1 m
L = 1000 m
L = 0.001 m
<< L
~ L
>> L
0
1e+08
2e+08
3e+08
4e+08
5e+08
6e+08
0
0.2 0.4 0.6 0.8 1
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Nondimensional film thickness
(b) Flux problems
L = 0.01 m
L = 0.1 m
L = 10 m
L = 1 m
L = 1000 m
L = 0.001 m
<< L
~ L
>> L
0
500
1000
1500
2000
2500
3000
3500
1e09 1e08 1e07 1e06 1e05 0.0001 0.001
T
h
e
r
m
a
l

c
o
n
d
u
c
t
i
v
i
t
y

(
W
/
m
/
K
)
Film thickness (m)
Film
Bulk
(c) Thermal conductivity
>> L ~ L << L
24.10 MULTILAYERS AND SUPERLATTICES
We also demonstrate applications relevant to the interface conditions in multilayers
and superlattices via the C- and F-processes model based on two schools of thought:
(1) the use of contact conduction and contact resistance of layered structures based
on geometric, mechanical load, and thermal aspects,
47
and (2) the consideration of
phonon transport in nanostructures.
48
In both cases, it is assumed that the interface
of dissimilar materials of the multilm structure has imperfect thermal contact at x
i
between layers (i 1) and (i). Via the present CF model, for method 1 the appropriate
interface conditions are given as
(24.56)
(24.57)
for i 1, 2, . . . , n. As applicable to the CF model,
47
Equation 24.56 represents
the continuity of the interface heat ux, and Equation 24.57 states that the difference
between the two surface temperatures is proportional to the heat ux. The propor-
tionality coefcient is the thermal contact resistance R
(i1)(i)
between the two layers,
which is a function of forces acting on layer (i 1) due to layer (i) and vice versa.
TABLE 24.2
Data Used to Predict the Relaxation Times of Diamond Type IIa
Films
Constant Properties
Bulk K
b
3320 W/m/K
Lattice constant 3.567 A
Specic heat 517.05 J/kg/K
Mass density 3510 kg/m
3
Phonon velocity
c
12,288 m/s
Constant A (Eq. 24.53) 188.06
a
Effective mean free path 0.447 m
b
(Eq. 24.51) 0.154 10
26
atoms/m
3
a (Eq. 24.53) 1.58
Debye temperature 1860 K
a
This value is slightly different from the one reported in Reference 6 because we have
chosen to compute all parameters based on the phonon velocity set to 12,288 m/s.
b
Anthony, T.R. et al., Phys. Rev. B, 42, 11041111, 1990.
c
Majumdar, A., J. Heat Transfer, 115, 716, 1993.
K
T
x
F
T
x
i
i
i
i
T
i
( )
( )
( )
( )
( )
1
1
1
1
1
1

]]
1

1
]
1
K
T
x
F
T
x
x
i
i
i
T
i
i
( )
( )
( ) ( )
( )
,

at
ii
K
T
x
F
T
x
i
i
i
T
i
i
( )
( )
( ) ( )
( )
1
1
1
1 1
1

]]
1

( )
+

1
1 1
R
T T
F
R
T
i i
i j
T
i i ( ) ( )
( ) ( )
( ) ( )
(

ii j
T
) ( )
( )

Usually, this is a parameter obtained by experiments and represents an unknown in
this formulation.
Alternatively, for method 2, via the CF model, the phonon transmission across
the interface can be accounted for based on the effect of diffusive scattering as
(24.58)
where
(i1)(i)
is the phonon transmissivity across the rough gap from the (i 1) to
the (i) layer. Method 2 formulation does not require direct knowledge of the thermal
boundary resistance, and after solving for the unknown temperatures it can be
obtained from Reference 48. As an illustration, GaAs/AlAs temperature proles
spanning space scales and the prediction of thermal conductivity are shown in Fig.
24.7. Thus far, we have demonstrated that the C- and F-processes model can (1)
explain the heat transport behavior from ballistic limit (Casimir limit: >> L) to the
diffusive limit (Fourier limit: << L) for the prediction of the thermal conductivity,
(2) span time scales explaining nite to innite speeds of heat propagation, and (3)
overall provide a balanced energy equation with the inclusion of a nondimensional
heat conduction model number, which is given as the ratio of the conductivity
associated with fast processes to the total conductivity composed of both fast and
slow processes. It is simple to implement and circumvents many of the deciencies
existing in the literature.
24.11 EQUATION OF PHONON RADIATIVE TRANSFER
(EPRT)
In thin dielectric lms, the major heat carriers are phonons. The thermodynamic
equilibrium distribution for these carriers follows the BoseEinstein distribution,
which is similar to the distribution of photons. Under this premise the equation of
phonon radiative transfer (EPRT) was originally described in Reference 49 based
on the correlation between the radiation theory (i.e., the equation of radiative transfer
[ERT]) and the transport theory in dielectric thin lms. The major assumption made
regarding solving the transport properties in very thin lms is that the problem can
be solved as a one-dimensional gray medium between black walls at specied
temperatures under radiative equilibrium. This condition arises when a dielectric
lm is sandwiched between two metallic lms. The following is a detailed derivation
of this theory.
The intensity of photons, that is, the radiation emitted in any direction by a wave
packet, is given by
24
(24.59)

( ) ( )
( )
( )
( ) ( )
(
i i
i
i
i
T
i
K
T
x
F
T

+

1
1
1
1 1

ii
i i
i
i
i
T
x
K
T
x
F
1
]
1

+
1
1
)
( ) ( )
( )
( )
( ) (

ii
i
i
T
x
x
)
( )
,
1
]
1
at
I x t v f x t D
p
( , , , , ) ( , ) ( , ) ( )

FIGURE 24.7 Representative C- and F-processes model results for GaAs/AlAs superlattices.
T
L
T
L
T
i
T
j
T
s
T
R
T
R
q
.
n
q
.
n
(i 1) (i)
xi
(a) Schematic diagram of a
two-layer film of thickness
L between dissimilar
ambient temperatures.
300
299.99
299.98
299.97
299.96
299.95
299.93
299.92
299.91
299.9
0 0.2 0.6 0.8 1
L(i 1) = L(i)
AlAs
GaAs
Period = 40 Angs
Period = 400 Angs
Period = 900 Angs
299.94
T
e
m
p
e
r
a
t
u
r
e

(
K
)
0.4
Nondimensional period length [ x/(L( i 1) + L( i ))]
Temperature plots spanning
space scales.
(b)
60
50
40
30
20
10
0
1.00E 09 1.00E 08 1.00E 07
1.00E 06
Period length (m)
T
h
e
r
m
a
l

c
o
n
d
u
c
t
i
v
i
t
y

(
W
/
m
/
K
)
Thermal conductivity spanning
space scales.
(c)
where the summation is over the photon polarizations, v(, ) is the velocity vector
in the direction of and within a unit solid angle, is the energy at which the
photons propagate, and D() is the density of states per unit volume. The frequency
dependency of the phonon intensity is eliminated by assuming that the lm medium
acts as a gray body.
For cases where >> L, the EPRT appears to model microscale aspects encom-
passing the ballistic to diffusive limits. In these cases, the thermal conductivity for
thin lms is obtained at steady state due to scattering of phonons from the boundaries,
which returns the system to thermodynamic equilibrium. This suggests that the heat
conduction by phonons is similar to that of photons and thus can be analyzed as a
radiative transfer problem.
26
Applying the analogy between photons and phonons and multiplying Equation
24.6 by v
x
D() and using the intensity denition (Equation 24.59), the equation
of phonon radiative transfer is obtained as
(24.60)
where the velocity in the direction of the phonon propagation is v
x
v, cos,
is the angle between the phonon propagation and the x-direction, and I
0
is the
equilibrium intensity.
At steady-state, Equation 24.60 reduces to
(24.61)
Majumdar
6
analyzed the EPRT for both acoustically thin ( >> L) and acoustically
thick ( << L) limits and showed that the EPRT does indeed provide results that
span both limiting cases where the Casimir and the Fourier laws are applicable.
24.12 CALLAWAYHOLLANDS MODEL
Heat conduction in solids is well understood to follow the kinetic formula
(24.62)
where K, the thermal conductivity of an ensemble of heat carriers, is given by the total
specic heat C, the average speed of the heat carries v
s
and the mean free path .
When the assumption that all phonons have the same energy and velocity is not
desired, it is possible to account for the phonon dispersion where C, v, and are
frequency dependent. Equation 24.62 can be modied by including the summation
over all phonon branches (one longitudinal and two transverse) and by integrating

I x
t
v
I x
x
I T x I x
x
( , ) ( , ) ( ( )) ( , )
0
v
I x
x
I x I T x

+
( , )
( , ) ( ( ))
0
K Cv
s
1
3

over the phonon spectrum width of each branch as presented in the early works of
Peierls in 1929,
(24.63)
The temperature-dependent thermal conductivities of silicon and germanium lms
can be modeled based on the approximated solutions to the Boltzmann transport
equation (BTE) where the frequency-dependent relaxation times represent the
phonon-scattering events. The most widely used model for determining the thermal
conductivity is a method rst proposed by Callaway,
4
which assumes that the N-
processes dominate the scattering and the total thermal conductivity can be expressed
as a sum of K K
1
+ K
2
.
The rst modied form of Callaways model, known as Hollands model,
5
assumes that K
2
0 and includes the three phonon polarizations in K
1
:
K
1
K
TO
+ K
TU
+ K
L
(24.64)
Note that Hollands and the modied CallawayHolland models are solutions of the
steady-state version of the BTE and can only provide the temperature-dependent
results of thermal conductivity and therefore only address the issue of microscopic
size, i.e., they can only determine thermal conductivity as a function of size.
24.12.1 COMPARISON OF HOLLAND MODEL AND EPRT RESULTS
Figure 24.8(a) depicts the experimental and analytical results obtained by Asheghi and
colleagues
42
for crystalline silicon layers 0.42, 0.83, and 1.6 m thick. In their study,
yet another modication to Callaways temperature-dependent thermal conductivity
model is employed, which considers the possibility of specular reection through
the surface. Their results show the peak in the conductivity for the thin lms to
occur at 70 K. The recommended bulk conductivity reaches a maximum of 5500
W/m/K at 30 K.
42
Figure 24.8(b) shows the results from the nite element implementation of the
steady-state EPRT compared with the experimental and analytical results given by
Reference 42. Although the results in Fig. 24.8(b) present values that are higher
than those seen in the results of Asheghi and colleagues, they provide a better t
of the data at low temperatures. The current study of the EPRT did not include
effects of grain boundary scattering. The addition of the grain boundary scattering
reduced the thermal conductivity by adding more sites for the thermal resistance. The
results obtained by the EPRT method seem more appropriately used with lms of the
1-m range. Note that at room temperature the mean free paths of silicon and
diamond lms are 0.0409 and 0.39 m, respectively. Hence, lms thinner than 1
m will present microscopic ballistic behavior at room temperature.
K C v d
p
1
3
( ) ( ) ( )
The results shown with the nite element method (FEM) are based on experi-
mental thermal conductivity data available from work by Anthony and colleagues
46
(Table 24.3).
24.13 MOLECULAR DYNAMICS
Molecular dynamics (MD) is a computer simulation technique in which the time
evolution of a set of interacting atoms is followed by integrating their classical equa-
tions of motion corresponding to the second law of classical mechanics formulated
by Sir Isaac Newton in 1687
(24.65)
FIGURE 24.8 Asheghi et al.s
42
work compared to results using FEM to solve the EPRT in
determining temperature-dependent thermal conductivity for thin silicon lms.
Single-crystal
silicon
Bulk
Films
Predictions
Film
thickness
1.60 m
0.83 m
0.42 m
Data
Size
effect
10
4
1000
100
10
5
10 100 350
Experimental data and analytical results following
Hollands method of crystalline silicon layers of
several thickness (from Asheghi et al., 1998
with grain boundary)
(a)
10000
1000
100
10
10 100 1000
1
T
h
e
r
m
a
l

c
o
n
d
u
c
t
i
v
i
t
y

(
W
/
m
/
K
)
Temperature (K)
L = .1e6 m
L = Bulk
L = .42e6 m
L = .83e6 m
L = 1.6e6 m
FEM results following the EPRT method for silicon
layers of several thickness (without grain boundary)
(b)

F r m a
i N i i
( )
where is the force acting on the atom i at a given time in a system containing N
atoms, m
i
is the atom mass, and is the atom acceleration given by
with as the atom position.
However, the motion of molecules is very complicated and the forces involved
are nonlinear and depend on the positions of the other atoms. Therefore, a different
form of the classical equation of motion is used and follows Hamiltons equation
where particles are described by their positions and momenta . The forces
on the particles are obtained from the gradient of the interatomic potential energy
surface U(r
1
, r
2
, . . . , r
N
) as a function of the positions of all the atoms:
(24.66)
where operates on the position r
i
of atom i. Any change in the potential energy
that results from a displacement of atom i contributes to the force acting on atom i.
In order to develop a molecular dynamics simulation, it is necessary to specify
both the atomic structures of the nanotubes where atoms are arranged at specic
positions and the interatomic potential, which, for carbon nanotubes, is based on the
TersoffBrenner
5052
functional form with parameters tted from experimental and
quantum mechanical data to reect good structural, mechanical, and thermal properties
of nanotubes. A pristine nanotube is illustrated in this work. After the initial conditions
and the interaction potential are dened, the equations of motion are numerically
integrated and solved using a predictorcorrector algorithm with xed time steps of
0.2 fs. The molecular dynamics simulation provides information at the microscopic
level via the positions and velocities of all atoms as a function of time,
TABLE 24.3
Parameters Used in Solving EPRT for
Silicon Film
Parameter Silicon
v (m/s) 6500
(m) 0.044
A 101.4
(W/m
2
/K
4
) 24.8
D
K 625
F
a
i

a
d r
dt
i
i
2
2
r
i
r

p mv

F U r r r
i r N
i
( , , . . . , )
1 2
r
i

r t v t
i i
( ), ( )
In order to obtain thermal conductivity results from the information provided at the
microscopic level, a nonequilibrium molecular dynamics simulation-based direct
method
12,13
is used. The direct method emulates the actual experimental situation by
setting two regions of hot and cold atoms.
24.13.1 PRISTINE NANOTUBE
The results for thermal conductivity shown here are for (10,10) single-wall carbon
nanotubes (SWNTs). The theoretical expectation for SWNTs at room temperature
(300 K) is either 2000 to 3000 W/m/K
12,14
or about 6000 W/m/K.
15
As mentioned
earlier, results reported for the peak value of thermal conductivity in Reference 15
are an order of magnitude larger than the results reported in other work
12,14
and in
recent experiments.
16
This is because an extremely small sample of carbon nanotubes
was used in that simulation, and extrapolation of the data was probably erroneous.
It is noteworthy to mention that in running a MD simulation it was observed that
the thermal conductivity is highly dependent on the size of the nanotube and the
temperature gradient chosen, as shown in Fig. 24.9. The temperature gradient used
for obtaining the thermal conductivity results shown in Fig. 24.9(b), is kept constant
at 2.03 K/nm, and the size of the nanotube is varied from 98.4 to 787.0 (1600 to
12,800 atoms). As the size of the nanotube is increased, the thermal conductivity
appears to asymptotically approach a constant value. This is an interesting nding,
which is in agreement with experimental results found in thin lms, where the
thermal conductivity is size dependent. As the lm size approaches a critical value,
the thermal conductivity equals that of its bulk counterpart. This concept is worthy
of further investigation because the idea of bulk nanotube size is still under
investigation. Note that, for a given thermal gradient, which can be determined from
the experimental conditions, the thermal conductivity as a function of length con-
verges to a constant value within 40 to 80 nm for a (10,10) SWNT.
Figure 24.9(c) shows the change in thermal conductivity as a function of the
temperature gradient for a nanotube under periodic boundary conditions with
3200 atoms. Based on the expected room temperature thermal conductivity provided
in Reference 14, it appears that it would be more appropriate to choose a temperature
difference between 40 and 60 K for pristine nanotubes with 3200 atoms, instead of
80 K as shown in the results of Fig. 24.9(b). This nding, along with the question of
what will happen to nanotubes of other sizes, is being further investigated.
An introduction that serves as a prelude to an improved understanding of nanoscale,
microscale, and macroscale thermal transport was presented. Since the early work
in the 1930s, there has been a quest to provide an improved scientic understanding
of size effects in thermal conductivity. More notably, with the advent of nano-
technology, the 1990s brought a surge of research interest and activity in the area
of thermal transport. Although limitations exist in many of the methods, it is time
to take a critical and in-depth look at the various investigations, questions, and
concerns raised in the literature. The present chapter highlights some of the recent
advances in thermal transport analysis and also provides some challenging solutions
FIGURE 24.9 Thermal conductivity of carbon nanotubes.
Typical carbon nanotube (picture
courtesy of the National Center of
Competence in Research (NCCR) 4).
(a)
0
2000
4000
6000
8000
10000
100 200 300 400 500 600 700 800
T
h
e
r
m
a
l

c
o
n
d
u
c
t
i
v
i
t
y

(
W
/
m
/
K
)
Nanotube length (Angstroms)
(b) Thermal conductivity versus
nanotube length.
0
2000
4000
6000
8000
10000
0 1 2 3 4 5 6 7
T
h
e
r
m
a
l

c
o
n
d
u
c
t
i
v
i
t
y

(
W
/
m
/
K
)
Temperature gradient (K/nm)
Thermal gradient study
Thermal conductivity versus
temperature gradient.
(c)
to various issues appearing in the literature. It is hoped that this chapter fosters future
research in the directions discussed.
PROBLEMS
1. Identify the various methods (classical and nonclassical) existing in the
literature for the prediction of thermal conductivity at scales ranging from
nanoscale to macroscale regimes.
2. Conduct a thorough literature review and point out the limitations cited
in the literature for the classical models. Verify these limitations using
examples in the literature. Conduct the same for nonclassical models.
3. Derive the C- and F-processes constitutive model and understand the
notion of spanning ballistic to diffusive limits and nite to innite speeds
of heat propagation. Since the method is not computationally intensive,
verify some numerical examples such as thin lms, superlattices, thermal
contact conduction/resistance, and the like.
4. Investigate methods that are not computationally intensive yet do not require
tting parameters. For example, look into coupling the C- and F-processes
model with variants that are less cumbersome to molecular dynamics to
circumvent potential limitations.
5. Develop multiscale methods suited for modeling and simulation of large-
scale applications on modern high-performance computing (HPC) platforms.
ACKNOWLEDGMENTS
The support for this work is in the form of computer grants from the Minnesota
Supercomputer Institute and in part by the Army High Performance Computing
Research Center under the auspices of the Department of the Army, Army Research
Laboratory (DAAD19-01-2-0014). The content does not necessarily reect the posi-
tion or the policy of the government, and no ofcial endorsement should be inferred.
Special thanks are due to X. Zhou for related technical discussions. Thanks are also
due to D. Srivastava for collaborations on the MD simulations. Special thanks are due
to Amit Jain for the preparation of the chapter.
REFERENCES
1. Casimir, H.B.G., Note on the conduction of heat in crystals, Physica, 5, 495500, 1938.
2. Lambropoulos, J.C., Jolly, M.R., Amsden, C.A., Gilman, S.E., Sinicropi, M. J.,
Diakomihalis, D., and Jacobs, S.D., Thermal conductivity of dielectric thin lms,
J. Appl. Phys., 66, 42304242, 1989.
3. Klemens, P.G., Proc. Roy. Soc. (London), A208, 108, 1951.
4. Callaway, J., Model of lattice thermal conductivity at low temperatures, Phys. Rev.,
113, 10461051, 1959.
5. Holland, M.G., Analysis of lattice thermal conductivity, Phys. Rev., 132, 24612471,
1963.
6. Majumdar, A., Microscale heat conduction in dielectric thin lms, ASME J. Heat
Transfer, 115, 716, 1993.
7. Chen, G. and Tien, C.L., Thermal conductivity of quantum well structures, J. Thermophys.
Heat Transfer, 7, 311318, 1993.
8. Chen, G., Ballistic-diffusive heat-conduction equations, Phys. Rev. Lett., 86,
22972300, 2001.
9. Goodson, K. and Flik, M.I., Microscale phonon transport in dielectrics and intrinsic
semiconductors, ASME, 227, 2936, 1992.
10. Mazumder, S. and Majumdar, A., Monte Carlo study of phonon transport in solid thin
lms including dispersion and polarization, ASME J. Heat Transfer, 123, 749759, 2001.
11. Peterson, R.B., Direct simulation of phonon-mediated heat transfer in a Debye crystal,
ASME J. Heat Transfer, 116, 815, 1994.
12. Osman, M.A. and Srivastava, D., Temperature dependence of the thermal conductivity
of single-wall carbon nanotubes, Nanotechnology, 12, 2124, 2001.
13. Muller-Plathe, F., A simple nonequilibrium molecular dynamics method for calcu-
lating the thermal conductivity, J. Chem. Phys., 106, 60826085, 1997.
14. Che, J., agin, T., and Goddard III, W., A., Thermal conductivity of carbon nanotubes,
Nanotechnology, 11, 6569, 2000.
15. Berber, S.K., Young-Kyun and Tomnek, D., Unusually high thermal conductivity of
carbon nanotubes, Phys. Rev. Lett., 84, 46134616, 2000.
16. Hone, J., Whitney, M., and Zettl, A., Thermal conductivity of single-walled carbon
nanotubes, Synthetic Metals, 103, 24982499, 1999.
17. Cattaneo, M. C., Sur une frome de lquation de la chaleur liminant le paradoxe
dune porpagation instantane, C. R. Acad. Sci., 247, 431, 1958.
18. Vernotte, P., Les paradoxes de la thorie continue de lquation de la chaleur, C. R.
Acad. Sci., 246, 3154, 1958.
19. Joshi, A.A., and Majumdar, A., Transient ballistic and diffusive phonon heat transport
in thin lms, J. Appl. Phys., 74, 3139, 1993.
20. Tien, C.L., Qiu, T.Q., and Norris, P.M., Microscale thermal phenomena in contem-
porary technology, J. Thermal Sci. Eng., 2, 111, 1994.
21. Anderson, C.V.D.R. and Tamma, K.K., An overview of advances in heat conduction
models and approaches for prediction of thermal conductivity in thin dielectric lms,
Int. J. Heat Fluid Flow, 14, 1265, 2003.
22. Zhigilei, L.V., MSE 524: Modeling in Materials Science, Spring 2002.http://www.
people.virginia.edu/lz2n/mse524/.
23. Klitsner, T., VanCleve, J.E., Fischer, H.E., and Pohl, R.O., Phonon radiative heat
transfer and surface scattering, Phys. Rev. B, 38, 75767598, 1988.
24. Vincenti, W.G., and Kruger, C.H., Introduction to Physical Gas Dynamics, Robert
Krieger, Ed., New York, 1977.
25. Kittel, C., Introduction to Solid State Physics, 6th ed. John Wiley & Sons, New York,
1996.
26. Ashcroft, N.W., and Mermin, N.D., Solid State Physics, W.B., Saunders, Philadelphia,
1976.
27. Lee, J.F., Sears, F.W., and Turcotte, D.L., Statistical Thermodynamics, 2nd ed.,
Addison-Wesley Publishing Co. Inc., Reading, MA, 1973.
28. Zhou, X., Tamma, K.K., and Anderson, C.V. D.R., On a new C-F processes heat
conduction constitutive model and the associated dynamic thermoelasticity, J. Ther-
mal Stresses, 24, 531564, 2001.
29. Joseph, D.D. and Preziosi, L., Heat waves, Rev. Mod. Phys., 61, 4173, 1989.
30. Fourier, J.B.J., Thorie Analytique De La Chaleur, Paris, 1822.
31. Bird, R. B., Armstrong, B., and Hassager, O., Dynamics of Polymeric Liquids, John
Wiley & Sons, 1977.
32. Joseph, D.D., Narain, A., and Riccius, O., Shear-wave speeds and elastic moduli for
different liquids, Part I: Theory, J. Fluid Mech., 171, 41, 1986.
33. Jeffreys, H., The Earth, University Press, Cambridge, U.K., 1929.
34. Anisimov, S.I., Kapeliovich, B.L., and Perelman, T.L., Femtosecond electronic heat-
transfer dynamics in thin gold lm, Sov. Phys. JETP, 39, 375, 1974.
35. Corkum, P.B., Brunel, F., and Sherman, N.K., Thermal response of metals to
ultrashort-pulse laser excitation, Phys. Rev. Lett., 61, 2886, 1988.
36. Joseph, D.D., and Preziosi, L., Addendum to the paper Heat Waves, Rev. Mod.
Phys., 62, 375, 1990.
37. Tamma, K.K., and Zhou, X., Macroscale and microscale thermal transport and thermo-
mechanical interactions: Some noteworthy perspectives, J. Thermal Stresses, 21, 405,
1998.
38. Ziman, J.M., Electrons and Phonons: The Theory of Transport Phenomena in Solids,
Claredon Press, Oxford, 1960.
39. Berman, R., Thermal Conduction in Solids, Clarendon Press, Oxford, 1978.
40. Graebner, J.E., Reiss, M.E., Seibles, L., Hartnett, T.M., Miller, R.P., and Robinson,
C.J., Phonon scattering in chemical-vapor-deposited diamond, Phys. Rev. B, 50,
37023713, 1994.
41. Asen-Palmer, M., Bartkowski, K., Gmelin, E., Cardona, M., Zhernov, A.P., Inyushkin,
A.V., Taldenkov, A., Ozhogin, V.I., Itoh, K.M., and Haller, E.E., Thermal conductivity
of germanium crystals with different isotopic compositions, Phys. Rev. B, 56,
94319447, 1997.
42. Asheghi, M., Touzelbaev, M.N., Goodson, K.E., Leung, Y.K., and Wong, S.S.,
Temperature-dependent thermal conductivity of single-crystal silicon layers in soil
substrates, Trans. ASME, 120, 3036, 1998.
43. Brorson, S.D., Fujimoto, J.G., and Ippen, E.P., Femtosecond electronic heat transfer
dynamics in thin gold lm, Phys. Rev. Lett., 59, 1962, 1987.
44. Qiu, T.Q., Juhasz, T., Suarez, C., Bron, W.E., and Tien, C.L., Femtosecond laser heating
of multi-layered metals II. Experiments, Intl. J. Heat Mass Trans., 37, 719, 1994.
45. Pells, G.P., and Shiga, M., The optical properties of copper and gold as a function
of temperature, J. Phys. C, 2, 1835, 1969.
46. Anthony, T.R., Banholzer, W.F., Fleischer, J.F., Wei, L., Kuo, P.K., Thomas, R.L.,
and Pryor, R.W., Thermal diffusivity of isotropically enriched 12c diamond, Phys.
Rev. B, 42, 11041111, 1990.
47. Blandford, G.E. and Tauchert, T.R., Nonlinear thermoelastic analysis of layered
structures, Finite Elements Anal. Design, 1, 271285, 1985.
48. Chen, G., Heat transport in the perpendicular direction of superlattices and periodic
thin-lm structures, ASME, 59, 1324, 1996.
49. Majumdar, A., Microscale heat conduction in dielectric thin lms, ASME, 184, 3342,
1991.
ical vapor deposition of siamond lms, Phys. Rev. B, 42, 94589471, 1990.
51. Tersoff, J., New emperical approach for the structure and energy of covalent systems,
Phys. Rev. B, 38, 69917000, 1988.
52. Tersoff, J., Empirical interatomic potential for silicon with improved elastic properties,
Phys. Rev. B, 38, 99029905, 1988.
53. Van Zegbroeck, B.J., Distribution Functions, 1996, http://ece-www.colorado.edu/~
bart/book/distrib.html.
54. Peierls, R.E., Zur theorie der galvanomagnetischen effekte, Zeitschrift fur physik, 53
(1929), 255.

Nanoengineering of Structural Functional and Smart Materials

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nanoengineering of Structural Functional and Smart Materials

Uploaded by

Copyright:

Available Formats

Nanoengineering of Structural,

Functional, and Smart Materials

) (see the inset

component exhibits a weak diameter dependence. The

features of A symmetry, and these components

0.142 nm for making this plot.

) side of the substrate is used as the back gate electrode,

(AMU 15), and CH

(AMU 16). The partial pressure of CH

, and both species

. During the first growth experiment the methane flow rate

species first decreases and then increases accordingly. Because of

, indicating fluorine in the coating film. The fluorine can come

7.9 9.4 9.1

13.4 Conclusions and Applications

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

. The adhesion strength between PS molecules

): 70200 nm D, 50100 micron L; nonconcentric nested tubes with walls angled

Copyright 2006 by Taylor & Francis

30B, a MMT clay from

-filled EPDM rubber). The TPSiV X1180 ther-

I.28E, Nanomer I.30E, PWG,

Nanotechnology Conference, Huntington Beach, CA,

= 0, and that the additional

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

Nanoindentation: load - displacement relationship

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

Copyright 2006 by Taylor & Francis

XP by the process explained in previous chapters.

Copyright 2006 by Taylor & Francis

units. The short-

units. The overall dimensions

units of the polyethylene chains

Conductivity ( ) associated with f fast F-Processes

Copyright 2006 by Taylor & Francis

You might also like